This newsletter is supported by Swan AI.

For over a year, I’ve followed Swan’s mission to build the first autonomous business - $10M ARR per employee with AI agents. In year one, 3 founders landed 200+ customers across 5 continents and $1.5M in monthly pipeline, with zero employees or SDRs. This year they raised $6M to build the first shared context and workflow systems purpose-built for AI agents.

The architecture matters more than the agents.

They just revealed how they structure it: the truths that repeat across your GTM - ICP, scoring, voice, routing - live in one place, and every workflow points to them. So when your ICP shifts, you fix it once instead of editing it across fifty plays and tools.

Get the workflows

(free, non-gated)

Dear GTM Strategist!

I often get asked, “We need a GTM slide for our pitch deck - can you please review this”?

Then I get something that more resembles a wishlist of GTM motions and a clusterf* of various marketing and sales channels that should or could work down the line.

And investors are pretty great at smelling insecurity. “All” you need to do is to make them believe that the team knows what they are doing and how they will get there.

How does the opposite of it show?

In a GTM slide that is full of vague goals, but shows little of what will actually drive the growth.

More than 2 years ago, I posted a guide to a killer GTM slide with an investor, Jonathan Crowder, and it became one of my best-indexed posts - but much has changed since.

More competition, AI native trailblazer startups setting bigger appetites with investors and in-built expectations that nearly every team should somehow use AI in their product and/or GTM - to name some. VC is now more selective, more AI-centric, and more concentrated in a smaller number of winners and mega rounds.

I’d immediately argue that this one is not useful just for startups fundraising but to all of us who are working on GTM presentations for clients, bosses and team members - investors can be internal or external :)

The sole mission of this article is to empower you with structural and tactical insights into how to make a compelling pitch to people that you need on your GTM journey.

The logic of the GTM Power Slide is simple - the purpose of the GTM slide is to create confidence and evidence that your GTM plan will work by providing traction data and projections that do the story for you - and it will be faster and better with their contribution, for which they will be handsomely rewarded. Not much different than sales, ain’t it?

This is why this article did not need a makeover - it needed a do-over.

We asked well-respected investors in the GTM space to share their best practices with us: Liz Christo from Stage 2 Capital, Paul Irving from GTMfund, Kyle Poyar from Growth Unhinged, and Ivan Landabasso from JME Ventures.

We got some really good insights and packed them into a “pitch perfect” GTM slide for 2026 that you can use as a template for your presentations.

So, let’s get into it.

What changed since the pre-AI native era?

The market split in two - and your GTM slide now decides which side you land on.

• AI took 80% of global VC in Q1 2026. Average AI deal sizes hit $51M versus $4.7M for non-AI - a 10.9x gap, up from 8.4x in 2024. Outside AI, capital is tight.

• Series A is a revenue test now. Investors want real revenue, not a demo, and they expect fast growth. Carta puts the median Series A post-money at $78.7M, and the top 10% of startups raised half of all 2025 capital.

• VCs are writing fewer, larger checks. Over 40% of seed and Series A dollars in 2026 went to rounds of $100M+. Everyone else faces a tighter, more scrutinized environment where financial discipline and a clear story matter more than ever. Your GTM slide is that story.

In a nutshell: investors are writing big checks to a select few while the rest have to live with scraps and “valuable feedback”. It is an especially painful problem on the periphery - outside San Francisco, London, and other VC power hubs.

How to convince them with the GTM slide?

Step 1: Show traction

We won’t focus on product-market fit, “problem-solution” slide and other parts of the presentation today. These come before and create the narrative that will make the GTM claims more believable.

And then there’s one slide where you can expect most discussion: your go-to-market plan. You need to nail it.

So more than anything, you need to show where you are starting from: your existing traction.

You need to give the confidence that you know what you’re doing. Your GTM slide should show a realistic picture of how you plan to acquire customers in your market, supported by clear assumptions. Show your existing traction and present a story on why this is an investment opportunity the investors should be part of.

Step 2: Outline a clear strategy to continue ahead

Here’s where the story part should continue. Anyone can draw a graph with projected growth. You need to explain what will drive it. Which GTM motions are bringing in revenue right now, and what are your bets that the investment will boost?

The most common mistake is to present too many options. Then it resembles a wish list, but no one will know what the focus should be. Instead, explain what your key drivers of growth are, why you make that assumption, and how you will measure it (what the goals and KPIs are).

Kyle Poyar (Growth Unhinged, previously partner at OpenView) explains: “VCs want to see that your GTM is repeatable, predictable, and efficient. You know your best customer, you have channels that are working, your sales reps are hitting quota, and you know exactly what you could do to accelerate growth even further.”

For the earliest stages (Pre-Seed & Seed), the GTM slide usually isn’t that deep - and it doesn’t need to be, says Paul Irving, general partner at GTMfund. Three things stand out to him:

1. An unfair advantage in distribution.

For example, a built-in audience, a media asset you own, strong advisors, or founders who came from the space. “It’s not a prerequisite, but it’s the first thing I look for, because it lets you iterate with customers faster than a competitor can.”

2. A motion that fits the business.

Where your customers are, how they buy, what kind of company you’re building. Too often it’s cookie-cutter and doesn’t match the business. For example, outbound email to restaurant owners who never open email. The strategy has to match the specific customer.

3. One to three channels you can own.

“I’d rather see one or two channels they’re going to win outright to get off the ground than a seven-channel distributed plan an early team can’t run,” says Paul.

Step 3: Let the numbers do the talking

Numbers are the foundation. It’s how the valuations are made and justified.

So, which metrics to focus on?

It depends on the round you are looking for:

First step (even if you’re not preparing your GTM slide yet): start measuring the numbers so you can then pull them together when needed. CRM is the best source of truth to do that.

Liz Christo (partner, Stage 2 Capital) says: “I see a lot of decks that are incredibly qualitative and subjective. They communicate how they are feeling during their sales process rather than reverting to the data. Often I think that’s because they don’t actually have the data. My big encouragement is whether you are a founder selling or have a large team, get the right CRM hygiene in place so you can track basic metrics: leads, opportunities, customers, ACV (annual contract value), go-live dates, etc...As the old adage goes, you can’t improve something you can’t measure.”

If you’re pre-revenue and want to focus on the traction, here’s what we have previously defined with Jonathan:

If you are already operating with revenue numbers, ARR (annual recurring revenue) is the gold standard. But for detailed talks about the valuation, investors will be interested in the economics behind the numbers. For example, how much does it cost you to acquire customers?

Kyle Poyar suggests that your gross margin adjusted CAC (customer acquisition cost) payback period should be healthy - ideally <12 months for SMB and <18 months for midmarket or Enterprise deals. But: “Make sure to zoom out; VCs aren’t looking for the details, they want to see that what you’re doing is ready to 10x.”

The killer GTM slide

The basics still hold. Your GTM slide has two jobs:

1. Communicate that you can predictably and successfully create traction.

2. To present the big opportunity you can pursue by securing funding.

Here’s what a template for the GTM slide in your pitch deck looks like in 2026:

Let’s break it down:

• Existing traction: a chart of your existing metrics: focus on the number of customers, ARR, and what GTM motions are driving growth. Explain key highlights of what’s working in a few bullet points. That could include learnings.

• Projected growth: continue with a projection of how this growth will look if you get an investment that will support expansion of your activities.

• Untapped opportunities: which are the additional opportunities that the investment will help you unlock.

“A company that is just getting started with founder sales and their first few customers should be focused on how they are going to get in front of their target audience, what they are learning, short-term milestones to prove value and product-market fit,” explains Liz Christo. On the other hand, a company with a team of 10 ramped sellers in seat should be sharing real data on the funnel, conversion, rep performance, and a believable path to scale that motion.

Where founders usually get it wrong

1) The math doesn’t add up

VCs talk in numbers every day, so the fastest way to lose them is numbers that don’t square. Paul Irving says: “ACV, customer count, revenue goal, and the pace to get there all have to line up, and often they don’t. Someone says they’ll hit $10M ARR; that’s 132 customers. They’re signing three a month — how do they get to twelve?”

Or the ACVs they’re projecting look nothing like the ones they have today. Sometimes there’s a real answer (they just turned on PLG onboarding, so velocity will climb). But often the numbers simply don’t square with the business.

2) Unrealistic targets

If you were on a first call with someone, this doesn’t mean pipeline yet. Liz Christo says: “I often see slides that make outlandish claims like “$15M in pipeline created last quarter” as a pre-seed company, and the company is layering in assumptions. For example, the first call with Bank of America enters pipeline at $3M, when in reality it probably isn’t even a qualified opportunity, let alone ready to have a $$ figure associated.”

Numbers also need to be put into context, explains Liz Christo. Random numbers on slide, like “Closed $300K in last 30 days” will raise questions on why that is important:

• Did you only start selling 30 days ago?

• Is this an acceleration?

• Is this ahead or behind your target?

3) Faking traction

Most VCs have a well-developed bullshit detector, says Ivan Landabasso (partner at JME Ventures and author of Startup Riders). “Don’t use the term “design partners” and other buzzwords when trying to build an argument around product-market fit or validation. You either have an early client who’s hopefully paying you (even if it’s a small amount), or a full-on client.”

If you have a “design partner” who’s not paying you after a few months, it will raise a question if:

a) they are taking advantage of you (if you are solving a real problem, there’s no reason not to pay you even if it’s a small amount), or

b) if you are solving a problem that is not important enough.

How to say “design partner” in non-jargon? “Temporary non-paying customer until we prove value”.

What to do this week

GTM slide is not an overnight task. Get it drafted, gather some feedback, test the narrative, practice the pitch. That’s the best bet for you to get it right and win investors in the end.

Also, there’s nothing wrong with showing weakness as well - frame them as opportunities when you give them the attention they deserve. For example: if you haven’t figured out paid ads yet, you might do it when you can devote some budget to it.

The same goes for learnings. Liz Christo shares an example: “You might have tested events last quarter, launched a new geo, or tested a new partnership - these are all opportunities to learn about what is/isn’t working. Experimenting and learning fast (even failures) are critical to growth. This behavior demonstrates the mindset of the founder or the founding team. I don’t need to believe you have it all figured out, but willingness to learn, being introspective, and keep getting better means you will find a path.”

And to wrap up, here are the highlights from the experts that contributed to this post:

Till next time,

Maja

✅ Need ready-to-use GTM assets and AI prompts? Get the 100-Step GTM Checklist with proven website templates, sales decks, landing pages, outbound sequences, LinkedIn post frameworks, email sequences, and 20+ workshops you can immediately run with your team.

📘 New to GTM? Learn fundamentals. Get my best-selling GTM Strategist book that helped 9,500+ companies to go to market with confidence - frameworks and online course included.

📈 My latest course: AI-Powered LinkedIn Growth System teaches the exact system I use to generate 7M+ impressions a year and 70% of my B2B pipeline.

🏅 Are you in charge of GTM and responsible for leading others? Grab the GTM Masterclass (6 hours of training, end-to-end GTM explained on examples, guided workshops) to get your team up and running in no time.

🤝 Want to work together? ⏩ Check out the options and let me know how we can join forces.

This newsletter is sponsored by 1mind.

Buyers tell AI things they’d never tell a salesperson. “Send me more information” usually means “goodbye now” - and AI reads it correctly, while most sellers keep chasing a prospect who already checked out.

1mind analyzed hundreds of thousands of real enterprise buyer-AI conversations. Three findings stood out:

→ How someone talks to your AI predicts their seniority, function, and deal value within a few turns

→ In-market demand is far more concentrated than the “5% are ready” rule implies - and it shows in what buyers say, not a form

→ How buyers talk to your AI mirrors how they prompt AI to find you - conversation design and AEO are the same surface

Full report drops on a panel hosted by my good friend Jonathan Kvarfordt: AI Conversations that Convert.

Tuesday, June 30 at 12 pm EST. I’ll be there.

Register here

Dear GTM Strategist,

You’ve probably been there.

5,000 target companies that could benefit from your solution. A sequence you wrote and rewrote until it sounded exactly right to you. You’ve done housekeeping too - selected the most praised tools that promised great returns, verified emails and did everything the playbooks tell you to do.

Yet, when you hit play, nothing happened. No one cared.

Could we blame:

• The subject line?

• The sending tool?

• The offer?

Not really - in a world where decision makers are bombarded with emails, DMs, and robocalls, it takes 2 seconds to make a decision whether a message is relevant to them. Whatever sounds like: “Hi, congrats on building the X company - you growth is impressive, but I realized you are completely neglecting Y - this is exactly why we build Z - to help companies like X grow faster. Worth a 15min chat? You can book a demo here.”... will be largely ignored.

So what makes your message relevant to someone right now?

There are two points of failure in most outbound campaigns we are receiving and analyzing:

1. You emailed the wrong people. Only 0.1% of regular cold email campaigns convert to a meeting (source). The list wasn’t matched to anyone in real pain. It was just big.

2. The second failure is the one nobody admits. Even when you did reach a real buyer, you had nothing to say to them. You pulled the same signals everyone else pulls - the funding round, the job change, the LinkedIn post. Your “personalized” email opened exactly like the other nine in their inbox that morning. 🥱 This is not real personalization.

Everyone has the same AI models, the same enrichment, the same signals. The context you feed the machine is the only real moat left. And the only way to grasp relevant context is to DYOR - do your own research. That is the only way you’ll have something relevant to say to your people. Attention is deserved - not granted.

Yes - this is a post about the good boring research layer. About the 80% of outbound nobody talks about, and the half that makes a real difference in campaigns.

What you will learn (in this order):

1. How to identify the right segment that cares about your solution now

2. How to get, enrich, and filter the data

3. How to score it and prepare for campaigns

4. How to write the message that makes the buyer think, “wait, how do they know this about my market?”

Claude Code will be our super-assistant all along the way.

Quick note on the “we.” I handle the GTM strategy. My partner Anže Voje and his team build these workflows in Claude Code and Clay, on live client campaigns. What follows is the method we run daily.

It starts with a question most teams never think to ask.

1) Identify the segment with the highest pain point

Before you build a single list, answer this: where does proof of your buyer’s pain actually show up? How can you constantly monitor for it?

Don’t start by prompting “Build me a list with VP of Sales at a 200-person SaaS company.” Our goal is to find VPs who are, right now, drowning - because drowning is the only condition under which a stranger picks up.

So, where does drowning show up?

Picture a company that posts 12 “Account Executive” openings in six weeks. This is a signal that a sales org is scaling faster than it can possibly ramp. And this is the exact moment when an enablement or onboarding tool flips from nice-to-have to must-have. So a job board will hand you these signals - and Claude Code is how you can do research and then source the right data.

Signals beat demographics every time, because a signal has a clock on it. And the clock matters more than people think: a funding round from 150 days ago is not the same signal as one from yesterday. Fresh pain converts. Act fast.

Data you pull from public databases needs further work. Expect 30 to 40% of any pull to be miscategorized - the wrong part of the value chain, consultants disguised as buyers, media companies, outdated headcount. That’s where the next step comes in.

2) Map where your signal lives

There are many sources when you can find good data online. Claude Code will help you strategize and select the right sources and, if you can get your hands dirty, also access the data.

In Claude Code, describe your product and ICP, then have it build and populate this matrix:

Here are some common source categories you should check:

• Public filings and registries - permits, licenses, grants, regulatory disclosures, procurement records

• Hiring data - job postings are among the strongest “active pain” signals; companies hire right before or right after the pain becomes real

• Industry databases and association directories - membership lists, certified-partner lists

• Product and review data - G2/Capterra, app-store changes, integrations added or removed

• Self-reported activity - the company’s own site, press releases, conference appearances and exhibitor lists

• Data vendors/aggregators - broad coverage, but noisy and often stale; best for firmographics, weak for triggers

3) Find the right contacts with Claude Code

That hiring surge tells you a company is in pain. It does not tell you who to email. The 12 postings are signed by “Talent Acquisition,” not by the VP who owns the number and the budget.

This pattern repeats in almost every market:

No single source holds both “who the buyer is” and “what they are doing right now”.

You always operate with two separate lists:

1. First, you build a list of target companies

2. Then, you look for target contacts within these companies.

To find the right target contacts, ask Claude Code for a person - ask it to work backwards from the company. You already have the domain (URL) from step 2. Now point it at where decision makers are actually listed.

Here’s an example of a prompt you can use:

For each company in this list, find the person who owns the sales number. Check the leadership or team page first, then LinkedIn. Look for VP of Sales, Head of Revenue, or CRO. If the company is under 50 people, assume the founder or CEO owns it. Return name, title, LinkedIn URL, and the source you used for each one.

Claude Code goes company by company - reading the leadership page, cross-checking the title on LinkedIn, and flagging where it had to infer versus confirm. That last part matters: you want to know which names are solid and which are educated guesses before you spend a send on them.

If you subscribe to any databases, like Apollo, connect them to Claude Code (via MCP - just type /mcp), and Claude will include these sources in research.

My go-to tool for research in Claude Code is Exa. Exa was built for AI agents, not humans, so it searches by meaning instead of keywords. You can hand it a messy query like “companies that posted 10+ AE openings last month and just raised a round” and it returns the actual page content, not ten links to scrape - so Claude can work with the output right away.

If you target larger companies, it’s often not just one person, but a buying group. So tell Claude Code to map the group:

For any company with over 200 employees, identify 2 to 3 people in the buying group: the economic buyer who signs off on the budget, the champion who feels this pain daily, and the user whose workflow your product touches. Note each person’s likely role in the decision.

This is exactly the work Claude Code is the superior tool for this type of work - it reasons across mismatched sources to construct the person that is not listed elsewhere.

When you get the list, you can ask Claude to export it as CSV, to Google Sheet… however you want it.

4) Now score it, or you’ll drown in your own list

So now you have your list and your contacts. The next step is to get ready for the campaign. You will be tempted to just send a “email blast” to the whole list. But - take a deep breath. You don’t want to burn all your chances immediately. And you want to be especially prepared for the top potentials and personalize the message to them.

That’s where lead scoring comes in. Don’t skip it.

Here is an aggregate example of a scoring table that we use in our campaigns:

Adjust it to your needs, but make sure to keep two components that make this lead scoring model very powerful

1. Signal stacking: a Series B plus a new RevOps hire in the same week is worth more than either alone, because together they tell you budget exists and rebuilding is underway. T

2. Penalties: Noisy lists overflow with things that pattern-match to your buyer and are categorically wrong - the manufacturer when you sell to the installer. Punish those hard and the list cleans itself.

Then sort into four buckets: HOT goes out now with your deepest personalization. These emails deserve your manual review and tweaking. Think of contact-specific resources and information you can mention. Can you provide a report or a demo for them that can give them a taste of how specifically you can help them? Do it!

With WARM contacts, you can proceed with a lighter touch. At COLD contacts, you watch for a trigger - the right timing to send. SKIP you archive so you never spend another credit on it. Here’s an example for a 14-point model.

Earn the reply

All this research buys you one thing: the right to be specific. And specificity is the whole game, because everyone is fishing the same pond with half-baked messages. The fix is to go deeper than the surface signal everyone shares, until your message carries something they cannot get anywhere else.

We have to deserve their attention by providing something valuable to them at the right time. Aim for a message that is valuable on its own, whether they buy or not. Market intelligence disguised as outreach. And remember - the first message should be all about them. Don’t pitch slap, don’t write extensively about you and your company. Just earn the reply and the right to continue the conversation.

The bar is simple: would this email be worth their time even if you were selling nothing? If yes, you have earned the reply. If no, you are probably being ignored.

The reaction you are hunting for is one sentence: “How do they know this about my market?”

That’s why Claude Code is the best for research, because you can combine your knowledge of the industry (your knowledge base) with specific about the contact - and use that information for personalization.

For the full picture, reference the trigger, the names, the number, and try to name their exact situation.

Three layers in every first touch:

• Lead with what you saw.

• Support it with the figure that’s true at their size.

• Ask a genuine question - not a meeting request wearing a trench coat.

Good: “Am I reading this right, or does your team already have this handled?”

Weak: “Got 15 minutes Thursday?”

The first one opens a conversation. The second one asks a stranger for a favor.

We went deep on this idea with Jordan Crawford, the king of the PVP framework: building campaigns your prospects would pay to receive - if you want the full version.

How this runs in Claude Code

We already shared some prompts and examples, but now that you know the order, let’s get even more specific on how it happens in Claude Code - and how to make it compound instead of starting from scratch every campaign.

First, give Claude Code somewhere to think from. A plain project folder it reads at the start of every session. This will be your GTM repository:

We already published a detailed guide on how to build your GTM repository, aka your GTM brain - find it here.

Next, fill this folder with all your knowledge: ICP definitions, messaging documents, product decks, signals… All the research you’ll further do with Claude Code should write back here.

Research: turn the ICP into signals. Research based on job titles and company size is not enough.

Ask Claude where the pain shows up. Example of a prompt:

Read the ICP definitions. List 10 observable, time-stamped signals that prove a company is drowning in unramped reps right now. For each, name the public source and how to spot it. No demographics - only signals with a clock on them.

You get a ranked signal list and a source map, saved to the repo.

Then source the list: Point these signals at the live source and let it pull, clean, and cross-reference in one pass:

Example:

Take the top 3 signals from the list you just built. For each one, search the public sources you mapped - use Exa for web research and any databases connected over MCP - and pull every company currently showing that signal. For each company, return: name, domain, the specific signal it fired, the source URL where you found it, and the date the signal appeared. Only keep companies that match our ICP - drop staffing agencies, consultancies, media companies, and anything in the wrong part of the value chain. Skip any signal older than 90 days. If a company shows more than one signal, list all of them on the same row. Write the results to companies.csv, one row per company.

Next, find the right contacts at these companies. We covered that at step 3. The goal is to get contacts.csv file with the actual contacts you can contact.

And finally, score these leads according to the model you set.

Here’s an example prompt based on the logic presented earlier in the article:

Score every contact in contacts.csv against scoring-rubric.md. Apply both the positive factors and the penalties. For each row, show the points awarded per factor, the running total, the tier (HOT 10-14, WARM 6-9, COLD 3-5, SKIP 2 or below - including any negative totals from penalties), and a one-sentence reason. Flag every factor you’re inferring versus confirming from a source. Sort by total, highest first, and write it back to scored-contacts.csv.

HOT is the only tier you personalize by hand, so Claude Code’s job here is to draft, not send. You feed it the same evidence you scored on, plus your repo, and have it build each message on the one specific thing you already know about that company.

Example:

For each HOT contact in scored-contacts.csv, draft a first-touch message. Open with the signal and source on their row - lead with what we saw, not who we are. Pull one proof point or benchmark from the repo that’s true at their company size. Close with a single genuine question, not a meeting request. Keep it under 90 words, no pitch. For each draft, note which signal and which proof point it’s built on so I can verify before sending.

Four rules to hold the line:

• Draft, don’t automate. Claude Code gets you to 80%. The top tier earns the last 20% you add by hand - the detail the model can’t see from a row in a CSV.

• One signal, deep. Build on the single sharpest thing you found, not five shallow data points stacked together. Depth is what reads as “they actually looked at us.”

• Pass the test before it sends. Would this email be worth their time even if you were selling nothing? If no, it’s not ready.

• The first message is all about them. No company backstory, no feature list. Earn the reply, then earn the right to continue.

Then, and only then, think about automation

By the time automation is even on the table, the hard part is behind you. But here’s the line we don’t cross: the first 50 messages to your hottest leads get reviewed and sent by hand. Every single one. Hand that step to a machine and you’re back to low reply rates this whole process was built to escape.

Everything below HOT is what you can automate. Once the thinking is done, WARM and COLD can orchestrate through Clay as the execution engine, or proceed directly with the sequencing tool like Instantly for email or HeyReach for LinkedIn DMs. But that build is a story for another post. The full Signal-to-Pipeline playbook walks through how to take this into Clay and run it at scale.

The takeaway

Outbound stopped paying for volume. It started paying for homework.

The research used to take an analyst a week- most companies skipped it enirely and treated outbound as trial and error.

Now Claude Code does it in an afternoon. And if you want it to compound instead of starting from zero every campaign, you put it somewhere permanent - a context repository Claude Code reads automatically, so this quarter’s research makes next quarter’s sharper.

Do the thinking first. Don’t skip the research. Discover the patterns relevant for your industry and ICP. Then run it at scale. That order is the difference between a list that books a meeting 0.1% of the time and a pipeline that compounds.

We packaged the full method - including the templates, the source catalog, and the pre-launch checklist - into one free-to-access playbook you can run on your next campaign.

Get the full workflow

Now tell us: does your outbound start with a signal, or with a list? Let’s discuss in the comments.

Maja & Anze

✅ Need ready-to-use GTM assets and AI prompts? Get the 100-Step GTM Checklist with proven website templates, sales decks, landing pages, outbound sequences, LinkedIn post frameworks, email sequences, and 20+ workshops you can immediately run with your team.

📘 New to GTM? Learn fundamentals. Get my best-selling GTM Strategist book that helped 9,500+ companies to go to market with confidence - frameworks and online course included.

📈 My latest course: AI-Powered LinkedIn Growth System teaches the exact system I use to generate 7M+ impressions a year and 70% of my B2B pipeline.

🏅 Are you in charge of GTM and responsible for leading others? Grab the GTM Masterclass (6 hours of training, end-to-end GTM explained on examples, guided workshops) to get your team up and running in no time.

🤝 Want to work together? ⏩ Check out the options and let me know how we can join forces.

This newsletter is supported by Attio, the AI CRM.

Attio just shipped the biggest update in 2026: completely new Workflows that don’t just automate your work. They do it. Workflows became AI agents.

For years, CRMs reminded you to do the work: update the record, chase the deal, send the follow-up. Attio flipped it. You build an agent once - your best rep’s playbook - and it runs across every account: enrichment, scoring, outbound, briefings, follow-ups. Same standard every time.

And if you can describe the play, Attio builds the workflow for you. Tell it your goal. Review. Run. Close the deal.

See what you can build

Dear GTM Strategist,

When I invite someone as a guest writer, it’s because someone has a way of looking at growth that I think every senior B2B operator on this list needs in their head.

Mark Roberge is that person.

He was the founding CRO at HubSpot, where he scaled revenue from $0 to IPO. He co-founded Stage 2 Capital, which I think of as the most interesting VC firm in B2B - it’s run and backed by 1000+ go-to-market operators from companies like Anthropic, Lovable, Databricks, Zoom, etc. He teaches scaling at Harvard Business School. He wrote The Sales Acceleration Formula and recently published The Science of Scaling. If there is a person on the planet who has watched more B2B companies try to scale, succeed, and crater, it is probably him.

I asked Mark to share something most senior B2B operators are quietly getting wrong right now. The answer landed harder than I expected.

His read: Great businesses are failing unnecessarily due to their scale strategy.. They use gut feel where they should be using operational performance metrics. Half scale too late and too slow. Half scale too early and too fast. The cost of getting it wrong - in either direction - is enormous.

Scale too late and a competitor takes the market.

Scale too early, and you accelerate a broken business into bankruptcy.

What you’ll get in this piece:

• Why product-market fit is measurable - and the three-variable definition Mark uses with every Stage 2 portfolio company

• What go-to-market fit actually is - and the leading indicators that tell you it’s working before unit economics confirm it

• The Stay-Go-Slow framework - the operating cadence Mark uses with boards to scale at the fastest pace the business can absorb without breaking

If you’re planning a 2026 expansion - new product, new market, new motion - read this twice. Then send it to your board.

Over to Mark.

The Second Act Stumble

Steve had a way of making growth sound inevitable.

He had doubled revenue from $27M to $54M in twelve months. The board wanted another lap. Another 2x. Another miracle. They were polite about it, of course, but expectations have a way of sounding like commands when they come from the people who write the checks.

Steve stared at the numbers. The inside sales team was expanding in the U.S. mid-market and could squeeze another $30M out of the current motion. But not $54M. There was a $24M hole staring back at him from the spreadsheet, and Steve hated holes.

The fix, in his mind, was obvious. Build a new product. Sell it to the existing customer base. Dangle it in front of prospects. Slap “platform” next to the company name and watch the deals pile up. He unveiled the plan. Heads nodded. The board was sold.

Engineers abandoned the roadmap. Marketing tore apart the website and rebuilt it around the launch. Sales enablement ran boot camps so every rep could pitch the new product in their sleep.

The customer conference was the coronation. Except it wasn’t.

A year later, the new product had sold 10% of the forecast. Bugs, friction, broken promises. The slick collateral fell flat on impact. Meanwhile, the original engine - the one that had powered the prior year - had been neglected. An entire quarter was burned chasing pipeline for a product that couldn’t carry the weight. By the time anyone admitted it, the company was too far behind to catch up.

Steve didn’t just miss his number. He missed the point.

In the early years, founders understand they shouldn’t commit to a revenue goal too soon. They use design partners. They iterate. They tolerate uncertainty.

Then the company hits scale, launches a new product or enters a new market, and somehow forgets all of it. We assume the playbook from act one will transfer to act two.

It rarely does.

Some institutional knowledge persists. Far less than we assume. When we expand, we have to restart the product-market fit and go-to-market fit process from scratch. The journey should be faster the second time. It is never instantaneous.

So what does that process actually look like?

Product-Market Fit Is Not a Feeling

Every year, I ask the same question - to MBA students at Harvard, to founders in Stage 2 portfolio reviews, to keynote audiences. What is product-market fit?

The answers are always all over the place. A revenue milestone. A customer count. Inbound flow. A “feeling.” I disagree with every one of those.

Product-market fit happens when customers continuously realize the value they were promised when they bought the product. The cleanest quantitative proxy for that is long-term customer retention. When a customer renews, they have effectively re-bought. They are voting with their dollars that the value showed up. Across most categories, the bar is annual retention above 90%.

There’s one problem with that definition. Customer retention is a lagging indicator. By the time you know the cohort retained, you are six to twelve months past the moment you needed the signal.

So the best operators define a leading indicator of retention, or LIR. Some people call it the “aha moment,” a phrase Chamath Palihapitiya popularized at Facebook. The point is the same: identify the specific, instrumentable behavior that predicts long-term retention, and chase it relentlessly.

There is no universal LIR. It’s unique for every business. However, it is useful to think through your LIR using the above framework, simplifying the decision down to three variable: Percentage (P), Event (E), and Timing (T).

P - Percentage. The share of new customers that must hit the leading indicator for you to claim PMF. If your market has lots of well-funded competitors and a real first-mover advantage, set P lower (like 50%) and move faster. If you operate in a niche with high failure tolerance and a strong moat, set P higher (80%) and “nail it before you scale it”.

E - Event. The actual behavior that signals value realization. This is the variable that matters most. A good E has four properties:

1. Objective. Either it happened or it didn’t. “Processed the first transaction” qualifies. “Customer sees value” does not.

2. Instrumentable. You have to be able to track it automatically, at scale.

3. Tied to customer success. “Reduced processing time by 10%” qualifies. “Attended the QBR” does not.

4. Aligned with your unique value proposition. The whole company - marketing, sales, customer success - will be steered by this metric. It needs to push them toward your differentiation, not somebody else’s.

Consider starting with E oriented around “setup”, evolving it to “engagement”, and ending at “ROI.” Early on, “processing transactions” is enough. Later, “processed 100 transactions.” Eventually, “achieved 10% productivity gain.” The metric matures with the maturity of the business.

T - Timing. How frequently you measure it. Daily lets you learn fast. But not all product use cases call for daily usage or smooth engagement. Usually, a monthly view is a timeline that generates consistent usage metrics with a fast enough learning cycle.

Once P, E, and T are set, instrument the LIR by acquisition cohort. Each row is the month a cohort was acquired. Each column is the percentage of that cohort hitting the LIR by month of tenure. You now have a monthly readout of whether new customers are realizing value - quarters before you’d see it in the retention number.

Two things matter most in that chart.

Recent cohorts beat older cohorts. Old cohorts were acquired before your latest product, onboarding, and positioning improvements. They aren’t representative anymore. Limit how far back the analysis goes.

The LIR cohort chart belongs in your board deck and your company-wide dashboard. If you measure something this important quarterly in a private spreadsheet, the org will not align around it.

That is how you stop calling product-market fit a feeling.

Go-to-Market Fit: Can You Acquire Customers Profitably?

Hitting PMF gives you confidence that if you add 100 customers in a segment next quarter, most of them will realize the value. However, you have not proven that you can acquire and serve them profitably.

I call this go-to-market fit, or GTMF. The ability to acquire and retain customers consistently and profitably.

Unit economics, rather than GAAP-based profitability, are used to evaluate GTMF, as unit economics isolate the business costs associated with scale. Commonly used unit economics metrics in B2B software are:

• Payback period < 12 months

• LTV / CAC > 3

• Magic Number > 1.0

However, just like retention in the prior section on measuring product-market fit, unit economics are also lagging indicators. Fortunately, extracting unit economics back to leading indicators is far more formulaic and consistent across businesses.

Take LTV/CAC > 3. The figure below illustrates how we can extract the desired unit economic output back into short-term operational inputs.

You now have a tree of operational metrics: cost per lead, lead-to-customer conversion, leads per rep, ACV, gross margin, and annual churn. Set a goal for each input, plug them into the formula, and confirm that the LTV/CAC output you want falls out the other side.

Then watch them weekly. If the actual line stays above the target line, you are on pace to achieve GTMF. If it doesn’t, you can see exactly which input is breaking - lead cost, conversion, capacity, or churn - and fix that one specifically rather than guessing.

The same logic works for payback period and Magic Number. Decompose, set targets, and watch the leading indicators.

This leading industry approach allows you to operate your business many quarters ahead of your competition.

Scale Is a Pace, Not an Event

With PMF and GTMF in place, we are ready to scale the new segment. However, scaling at the wrong speed is an even more prevelant failure cause that the wrong time..

The classic mistake is to treat scaling as a single event that follows a fundraiser or an annual planning cycle. Set the annual number. Calculate how many salespeople are needed to hit the number. Hire all of those salespeople as quickly as possible, ideally in the next two weeks.

This strategy fails due to a lack of appreciation for the operational processes required to hire, ramp, and manage this new velocity of salespeople. Six months later, the majority of the new hires are under-performing, LIR and LIUE deteriorates, and the next board meeting is rough.

Scale is a pace, not an event.

Plan to add 10 salespeople a month for 12 months, not 120 salespeople next month. If the system absorbs that, accelerate to twenty. If it strains, pause, diagnose, fix, resume. The optimal pace is the fastest one that doesn’t break product-market fit and go-to-market fit. The leading indicator dashboards established in the prior sections serve as your speedometer.

At Stage 2 Capital, we use a quarterly operating motion called Stay-Go-Slow. We still build an annual plan. We also pre-agree on three questions with the executive team and the board:

• GO. What would we see by the end of next quarter that justifies going faster - and increasing the hire rate?

• SLOW. What would we see that justifies slowing down - and pausing hiring until the issue is diagnosed?

• STAY. What would we see that validates the current plan?

Each quarter, we check in. We evaluate where we landed. We adjust the next four quarters of hiring. We update the forward plan.

Most teams operationalize Stay-Go-Slow against three thresholds:

All three green: GO. Accelerate hiring. Update the forward plan up.

Mixed green and yellow: STAY. Hold the plan.

Any red: SLOW. Decelerate or stop hiring. Diagnose the issue. Fix it as soon as possible, hopefully in days and not quarters. Re-establish the hiring pace once the issue is addressed. Update the forward plan down.

Two things tend to come up the first time I introduce this with a board.

The first is the IRR worry. Investors get nervous that slowing down kills their return profile. It usually doesn’t. Successful exits happen on a five-to-ten-year arc. Acquirers don’t care about your Q2-to-Q3 growth rate from eight years ago. They care about the most recent two years, at most. The same lens is true for the next growth investor. Their perception of the scale capability of the business is weighted toward recent quarters versus 6 quarters ago. Disciplined pacing protects the long compound. Reckless pacing destroys it.

The second is the accountability worry. If you adjust the forward plan when conditions change, do reps stop hitting their numbers? Two answers. First, you keep tracking the original plan for performance management and comp. Otherwise, teams hide behind the revised forecast. Second, this isn’t about reducing accountability. It’s about managing the business with the latest information. Those are different things.

The Real Goal

Half the executives I meet scale too early and too fast. Half scale too late and too slow.

The risk of moving too slowly is just as real as moving too quickly. Competitors raise capital. Market windows close. Press releases pile up around someone else’s name. Movement matters.

But movement toward the wrong milestone is just expensive motion. Move quickly toward the right milestone, at the pace your organization can absorb. Instrument product-market fit and go-to-market fit. Then scale at the fastest pace those two instruments will tolerate.

The goal is not a short-term “triple, triple, double, double.” The goal is a long-term home run.

Good luck with the science of scaling.

Maja here.

If you read this and felt the small twist in your stomach that says we are doing this wrong - good. That is the most useful response you can have to Mark’s framework.

The most concrete thing you can do this week:

1. Write down your current LIR. Sit with your CRO or head of growth. Pick a P, an E, and a T. Don’t make it perfect. Pick something, instrument it, and revisit in 30 days.

2. Decompose the unit economic target your board cares about most - LTV/CAC, payback, or Magic Number - into the four to six input metrics that drive it. Set a goal for each. That’s your LIUE dashboard.

3. Run a Stay-Go-Slow conversation at your next board meeting. Even if you don’t change the plan, the conversation will surface where your team and your board actually disagree about pace.

This piece is based on Mark’s book The Science of Scaling. If you’re serious about scaling, I definitely recommend reading it, as it includes much more on using data to decide when and how fast to scale revenue. Mark is also donating 100% of the proceeds of the book to mental health, so get some copies for your team as well.

And for more GTM wisdom from Mark and his team, subscribe to the GTM in Practice Substack.

Until next time,

Maja

✅ Need ready-to-use GTM assets and AI prompts? Get the 100-Step GTM Checklist with proven website templates, sales decks, landing pages, outbound sequences, LinkedIn post frameworks, email sequences, and 20+ workshops you can immediately run with your team.

📘 New to GTM? Learn fundamentals. Get my best-selling GTM Strategist book that helped 9,500+ companies to go to market with confidence - frameworks and online course included.

📈 My latest course: AI-Powered LinkedIn Growth System teaches the exact system I use to generate 7M+ impressions a year and 70% of my B2B pipeline.

🏅 Are you in charge of GTM and responsible for leading others? Grab the GTM Masterclass (6 hours of training, end-to-end GTM explained on examples, guided workshops) to get your team up and running in no time.

🤝 Want to work together? ⏩ Check out the options and let me know how we can join forces.

This newsletter is sponsored by Miro - where my frameworks actually come to life.

I’ve been a Miro power user since 2020 - workshops, planning, prototyping, inventing new models, you name it, I’ve got a board for it. This week, I ran my GTM Blueprint workshop live at Canvas 26 in London, and honestly, this is what Miro does best: an AI-powered workspace where teams come together to do the thinking.

Run these frameworks yourself, test Miro’s AI capabilities, and find the templates that turn you into a creativity powerhouse.

Get the free templates

Dear GTM Strategist,

Most of the go-to-market “strategies” landing in my inbox lately are AI slop.

You know it, I know it, so let’s not pretend otherwise.

So this time I want to bring the conversation back to something material and timeless: the four models I run with every single consulting client.

I invented them, perfected them over five years, and stress-tested them on AI companies at every stage you can think of - launching a product, expanding into a new market, even crafting that first pre-PMF plan. More than 12,000 companies have used them so far. That’s 15 years of work distilled into a handful of frameworks - but who’s counting? :)

Here’s what I love most about them: they are collaborative. Every single one is built as a workshop because I still believe that real human brains, in a room, on a board, together, create something beautiful that no model can spit out on its own.

And they age really well. They still deliver tons of value when we’re developing our go-to-market plans with teams. I revisit them yearly for strategic planning in my business practice as well.

Why does this matter more than ever?

Products have never been easier to build or copy. AI demos overpromise and underdeliver, so buyers walk in more skeptical than they were even a year ago. In that world, your edge isn’t more output. It’s focus and proprietary insight.

These four models are how I take teams from GTM overwhelm to GTM-focused execution, fast. The templates are open-source and free (video tutorials included).

Here’s what you’ll walk away with:

• A 60-minute diagnostic GTM Power Hour to map your whole motion and ship a v1 of your GTM plan

• The Market Problem Map to get you from “my product can help everyone” to one testable ICP

• A 5-stage Differentiated Positioning workflow so you stop sounding like every other “vanilla AI tool” that makes your target customer yawn 🥱

• My best seller now on AI steroids: GTM Blueprint. A reverse-engineered, prioritized quarterly plan you can actually execute

• The exact free templates, and where AI belongs in each step

Bon appétit 🍽️.

1. GTM Power Hour: Your first ballpark plan in 60 minutes

I invented this one in Belgrade during my Swiss Entrepreneurial Program, which was packed with back-to-back mentorship sessions. When I travel, I want to make the most of my time, but six 60-minute sessions in a day is heavy on context switching, even for me. So I sat down and thought hard about what would actually move the needle for entrepreneurs in the short window we had together.

After working on hundreds of GTM strategies, the pattern was obvious: founders bump into the exact same problems again and again (ballpark 90% of new products have ICP and positioning issues that need fixing before we turn on the marketing and sales megaphones).

But founders hate to hear that as a “go back and do your research homework” recommendation- what they need is an action plan, not another pile of “could-dos” that sinks them deeper into analysis paralysis. So I mapped an outline to help them spot the weak points in their current GTM and pinpoint the next opportunities to grab.

That’s how GTM Power Hour was born, and people love how simple and actionable it is. No heavy prior analysis required, so it works even when you’re just getting started. It helps you map the key elements of your go-to-market motion and ship version one of a plan to hit the traction that’s mission-critical for reaching your next stage.

One rule: constrain it to the next 3 to 6 months. Most people genuinely struggle to plan further out, and our landscape shifts so fast that half your assumptions go stale the moment the next AI model drops.

If your GTM plan reaches much further than six months out, it’s probably fiction. Plan for the next two quarters, then let reality (and the next model release) update you.

The output is a document you can hand to investors, colleagues, or partners. This one exercise pulls everything swirling around in your head into a clean, presentable structure.

You can get it here, with a video tutorial to walk you through it.

If you are already further ahead, share it with someone who needs this clarity. 🫶

2. Stop selling to “everyone”: Focus 🎯 - pick your beachhead with the Market Problem Map

Spray-and-pray (or its respectable-sounding cousin, “diversification strategy”) is a dangerous game in GTM. The number one mistake you can make is stretching yourself too thin. For most of us, GTM isn’t even the sole focus - we’re also wrangling finance, leading the team, shipping the product, and personal life, to name a few. The last thing you want is your attention and resources scattered across multiple bets.

If your AI BI tool “helps solo entrepreneurs, but also enterprises, and hey, why not the public sector too - and developers, yes, them too, because who doesn’t want to make data-driven decisions?! 😏” you are going to have a brutal time accumulating a critical mass of proof that your solution actually works in any one segment.

Why does that matter more than ever? Whoever you want to sell to wants to know who got results before them, and those reference customers need to be relatable in size, industry, and use case. Sure, radical innovations can catch the wings of virality by making innovators and early adopters fall in love. But most of us (sorry) aren’t building something differentiated enough to bet the company on “users are gonna love it, they won’t be able to shut up about it.”

Beachhead to the rescue. Your ECP (early customer profile) should not be “finance people in startups and scaleups” or “anyone with a budget,” but the specific, winnable slice of the market where your solution creates obvious, repeatable value. The beachhead segment.

A well-defined beachhead segment hits four criteria:

1. A pressing need for your solution (high pain point)

2. Adequate willingness to pay (budget already exists)

3. The ability to produce case studies, testimonials, and references that help you move into bigger adjacent segments (healthy markets, reasonable TAM)

4. Accessibility - you can realistically reach it from where you sit today

Usually there’s more than one beachhead hypothesis worth testing before you bet the next three to six months of GTM on winning it. Most teams come up with three to five ICP hypotheses, and it’s extremely hard to validate all of them at once. So we test gradually, and the Market Problem Map is the best tool I know for choosing which bet to test first.

How it works: you map the relevant industry problems you solve, the pain points, and how they manifest across your segments. Then you score on three criteria - level of pain, ease of reaching them, and ease of implementation. On the other side, you get either your best segment, or sometimes even the best problem to address across several segments.

This is the exercise that takes you from “who could we possibly help?” to “who do we test first?” - and you can start today.

That output is gold for early positioning and messaging, and everything around it: LinkedIn posts, outreach emails, sales decks, landing pages, all of it. And yes, the result will shift over time as your product matures and you learn more about your market. That’s expected. Iterate with it rather than trying to nail it on the first try.

3. Positioning: the 5-stage workflow that kills the “vanilla AI tool” vibe

Now that we roughly know who we’re talking to and what problems we’re solving, we can sharpen our communication with a legit positioning exercise.

Fair warning: this is usually the toughest one, especially for technical teams. Marketers have a slight head start here because we’ve been trained to think this way. But the real magic happens in the collision of both worlds, where you come up with something that absolutely does not sound like AI slop or another me-too AI tool.

I built this workflow by combining models from three people I learn from: Andrej Peršolja, Emma Stratton (her book Make It Punchy is excellent), and, believe it or not, Alex Hormozi - his value-prop formula is my cheat sheet, I use it constantly.

It runs in five stages:

1. Map the competitive alternatives. What are people actually using instead of you right now to fix this problem? Then map the evaluation criteria your ICP uses to decide whether something is the right solution.

2. Score how satisfied your ICP is with those alternatives. This shows you where you have the best shot to win. Usually you’ll find one or two criteria you can genuinely compete on, or an alternative people are so unhappy with that you can flat-out beat it (assuming you’ve got the capacity to solve it better).

3. Map your competitive assets. Careful here. “Great customer support” and “amazing team” are things anyone can claim, and that you should have anyway. Filter those out. What’s left, what’s genuinely and radically true for you, those are your real differentiators.

4. Translate differentiators into benefits and proof points. Proof points are increasingly non-negotiable. People have been burned by AI hype and enter the buying process far more skeptically than they were a year or two ago. Backing every claim is both smart and fair to your audience. Good business ethics.

5. Land your first version of positioning: a value proposition plus three benefits and features. This is exactly what feeds your landing page, your sales deck, and the context documents you’d drop into a Claude Code or Claude Cowork GTM setup.

The language always needs fine-tuning, and that’s where human plus AI iteration earns its keep, over multiple rounds, until it lands in the hearts and minds of your ICP.

Full honesty: This is a simplified scheme of the process, but in practice it’s a set of exercises that need to be adapted to the industry and GTM stage. You can find more examples (together with the templates and prompts for sales decks and landing pages) inside my GTM Checklist and GTM Bootcamp - links below.

4. GTM Blueprint 🛣️: reverse-engineer your next quarter of traction based on your objective

You’ve done the thinking. Now it’s time to bring in some traction - and oh boy, do I have a good one for this.

My starting move is always reverse engineering. Start with one measurable GTM objective, grounded in the next three to six months (locking plans in stone for 12 months is risky - at this stage we don’t know what we don’t know, and markets shift constantly). Then feed it two things:

1. The outside view - research on competitors, the market, and industry best practices.

2. The inside view - your team’s unique insights and experience.

That combination produces a proprietary, intelligent GTM plan instead of generic AI slop.

The winning move is to make your inputs actionable and testable. “Do a LinkedIn post” or “influencer collab” is directional, but not actionable. So prompt your team (and your AIs) in a way that yields ideas you can actually run and measure.

Say the objective is 500 new users by the end of next quarter. Your ideas might be:

• A PVP (personalized value prop) campaign in Clay and Claude Code. Good old outbound, done intelligently enough to actually convert.

• A performance campaign with an industry-relevant influencer who already has the audience and traction - assuming you nail the audience-influencer fit.

• Squeeze more value from existing users - growth loops for sharing, or even a referral program.

I love ideation. Generating all these exciting ideas sparks so much joy. But here’s the trap: you can do anything, but you can’t do everything. We all operate with limited resources, so the next move is non-negotiable - a reality check on who does this work and what will actually move the needle.

That reality check is prioritization, and I run it with a dead-simple matrix: value vs. effort.

• Value: If this works, how likely is it to move me toward my objective?

• Effort: How expensive, slow, or hard is this to pull off?

🪄 Pro workshop tip: I hate to admit it, but AI can be far more objective than humans at prioritization. I’ll always value gut feel and human oversight, but AI doesn’t “fall in love with its own ideas” or push something just because it’s really, really awesome. So I prioritize with AI first, then revise with the team - it saves time and a lot of unpleasantness.

Once you’ve prioritized, you’re ready to commit. Present the plan as a roadmap or a downloadable doc - whatever is most visible and actionable for your team. Avoid slides: they’re too static and never get updated IRL.

I packaged this whole workshop as a Miro Blueprint, free to use. Take it for a spin before the summer out-of-office auto-reply purgatory sets in.

Where to start (and where to go deeper)

My one piece of advice on where to begin: start where your customer journey is most broken right now.

No clear segment yet? Run the Market Problem Map.

Clarity but no traction? Jump straight to the GTM Blueprint.

You’ve got four frameworks and enough to run one this week - that’s real, and it’s yours for free.

If you want the full systems behind them (the templates, prompts, and worked examples that make them sing), they live in my GTM Checklist. Grabbing it is also how you keep this work going - which means more free frameworks and tutorials landing in your inbox.

Let’s go to market!

Love 💛

Maja

✅ Need ready-to-use GTM assets and AI prompts? Get the 100-Step GTM Checklist with proven website templates, sales decks, landing pages, outbound sequences, LinkedIn post frameworks, email sequences, and 20+ workshops you can immediately run with your team.

📘 New to GTM? Learn fundamentals. Get my best-selling GTM Strategist book that helped 9,500+ companies to go to market with confidence - frameworks and online course included.

📈 My latest course: AI-Powered LinkedIn Growth System teaches the exact system I use to generate 7M+ impressions a year and 70% of my B2B pipeline.

🏅 Are you in charge of GTM and responsible for leading others? Grab the GTM Masterclass (6 hours of training, end-to-end GTM explained on examples, guided workshops) to get your team up and running in no time.

🤝 Want to work together? ⏩ Check out the options and let me know how we can join forces.

This post is sponsored by Ahrefs - a leading SEO & marketing intelligence platform that helps you accelerate the growth of organic traffic to your website.

Ahrefs just launched Agent A - an AI agent that builds and runs your marketing tools, reports, and automations based on your Ahrefs data.

Take it for a spin:

Try Agent A

Dear GTM Strategist!

We’ve been talking about AEO (Answer Engine Optimization) for over a year now, and honestly? Still a lot of talk, very little action.

The case for acting is obvious. Your customer has changed how they buy. AI search brings more qualified traffic that converts better. And for once, newer brands get a fair fight - because AIs don’t just reward the decades of SEO assets that make the giants nearly impossible to beat.

But here’s the question nobody can answer cleanly: how do you actually get AIs to take you seriously when your ICP types “What’s the best CRM for a Series B startup” or “compare Attio vs HubSpot”?

This isn’t a job for the SEO nerds and demand gen wizards anymore. The customer journey changed, your website traffic-to-conversion math changed with it, and that makes it a GTM problem - your problem.

And if you have no idea what to do about it, you’re in very good company. George Chasiotis, founder of Minuttia, asked 599 marketers about their AEO strategy. Only 14% felt very confident about it. The rest are running experiments they can’t validate, copying playbooks they’re not sure work, and watching organic traffic move in ways they can’t explain.

That gap between “this clearly matters” and “we know what to do” is exactly where most B2B companies are stuck right now. So I asked George to fix that.
In this edition, you’ll get:

• Where most teams get stuck with AEO, and why

• The 6 pillars to win AI search visibility

• Real examples of what works and what backfires, including a case study

• A practical way to measure AI search performance when the data is messy and the goalposts keep moving

Let’s hear it from George.

The State of AEO: What 599 Marketers Revealed

My team at Minuttia ran this survey with Kevin Indig and his Growth Memo newsletter. You already know the gist from the intro: lots of activity, not much confidence. So let me skip the “AEO is here” sermon and get into the numbers that should actually change what you do on Monday.

Almost everyone’s in - but most are just dabbling

82% say they’ve adopted AEO “to some degree.” That headline sounds intimidating until you look at the breakdown. Only 7.7% have a dedicated AEO team. Just 21.5% are actively implementing a real strategy. The biggest chunk, 25.9%, are running early experiments, which is a polite way of saying “poking at it.”

So when someone tells you everyone’s doing AEO, they’re technically right and practically wrong. Most “adoption” is tentative. If you build a committed motion with real ownership now, you’re not catching up to the pack - you’re ahead of most of it. That window won’t stay open long, though. 13% already call AEO their top priority, up sharply from a year ago.

Why nobody feels confident (and why that’s good news for you)

The intro gave you the 14% confident stat. Here’s what’s eating the other 86% - and notice that every single cause is something you can fix:

Measurement is the number one complaint. Not because tools don’t exist - we counted roughly 240 AI search tracking tools in a December 2025 study. The tools are everywhere; the good ones are rare. People can’t tell which numbers to trust.

18% can’t tie AEO to business value. No attribution means no ROI story, which means no budget when you go ask for it.

24.2% flat-out don’t know what works. They’re collecting playbooks and guessing.

Most companies are confusing being busy with AEO for being good at it. Doing the work and knowing the work is paying off are completely different things.

The takeaway is almost embarrassingly simple: fix measurement before you scale anything else. The team that cracks attribution wins the budget arguments and pulls ahead while competitors keep running experiments they can’t read.

Don’t just dump this on your SEO team

Most orgs hand AEO to whoever already does SEO. About 20% spread it across cross-functional teams, and a tiny 2.7% bring in outside help.

Defaulting to the SEO team feels natural, but think about what you’re actually doing: piling a brand-new, genuinely uncertain discipline onto people who already have a full plate, usually without new tools or a clear mandate. Then everyone’s surprised when there’s plenty of activity and no movement. Pick an owner on purpose, and give them what they need to actually win.

The traffic drop hiding under all of this

Nearly half - 47.2% - reported organic traffic falling. For 20.5% it was a moderate dip of 10-20%, and for 26.7% it was a serious drop of more than 20%.

Fair warning: it’s not clear how cleanly people separated “AI search ate my traffic” from everything else that moves organic numbers, so treat this as directional rather than gospel. But you don’t need a perfect attribution model to feel the floor shifting. Waiting until the data is airtight is itself a choice, and probably the wrong one.

Bottom line: this space is still messy and young, and that’s precisely why there’s room to win. Let’s get into how.

6 Pillars of AI Search

The problem with how some organizations approach AEO is that they’re trying to reinvent the wheel.

That’s why so many of them are ready to throw SEO out of the window, declaring AI search the new sheriff in town.

But search is still here, and it’s more relevant than ever.

Many of the best practices you used for search engine optimization are still relevant, but we can certainly say they have evolved.

Take a look at the following graph, which provides an oversimplified view of what SEO used to be about:

At Minuttia, we still rely on these six pillars to help our clients with AI search visibility, but with a few tweaks.

Let’s take a closer look at each of them.

Pillar #1: Topic breadth and depth

These two notions are interconnected.

Topic breadth encompasses all topics relevant to you and your business model.

Let’s say you’re a CRM company, and that’s the core topic you’re talking about, but there are also the related ones:

• sales pipeline

• customer support

• marketing automation

• and so on.

In its turn, topic depth describes a set of sub-topics and the questions people ask for each of the main topics.

Let us illustrate.

Ideally, you need both topic breadth and depth to lay the foundation for successful AEO and SEO strategies, but there’s a caveat:

You should keep these topics as close to your business model as possible. Let us show you what can happen if you stray too far from it.

Connecteam, a workforce management software company, created a clever (at the time) strategy to boost Google and AI search visibility.

They created a repository of review articles analyzing 500+ of their competitors (as of January 2026).

To add weight to these reviews and show authority to AI search engines, Connecteam created a review methodology page explaining why readers should trust them.

You can find an entire breakdown of Connecteam’s AEO cast study here.

However, right after Google rolled out its Core Update in December 2025, which flagged content unrelated to a website’s area of expertise, Connecteam’s experiment started going downhill.

Organic traffic was the first one to take a hit, but AI citations also followed.

Why did that happen?

The nature of these pages goes against what Connecteam is - a workforce management tool, not a review site, like G2.

Besides, 500+ pages isn’t a small number, so their impact on Connecteam’s visibility is real.

We’re not saying you shouldn’t give up on comparative and self-promotional content.

They simply shouldn’t constitute such a big portion of your content.

Connecteam posting 500+ review pages and listicles, sometimes 16 in one month, is a perfect segway into the next point: Does publishing velocity impact your AI visibility?

Pillar #2: Publishing velocity

Now, it’s well-known that publishing velocity impacts organic visibility.

But we believe it has as much impact on AI visibility.

You already know that Google gives preference to expert content related to the website’s subject matter (Google has doubled down on it lately, with its December 2025 and February 2026 updates).

Data shows that AI search engines behave the same. According to Ahrefs, the average age of content cited by AI assistants is 1064 days - 25.7% “younger” than content cited by Google, which is 1432 days old on average.

What’s more, Ahrefs’ study proves that ChatGPT and Perplexity order citations from newer to older.

We know what you’re thinking: 1,064 days is basically 2.9 years. In what universe is that ‘fresh’?

Sure, both Google and AI assistants seem to cite long-lived content that has passed the test of time. However, your new pages will reach that moment in time once, too; that’s why you should keep them coming, but don’t sacrifice quality for speed.

On that note, as your content inventory keeps growing, you’ll need to find a way to strike a balance between publishing new content and updating older pieces, which leads us to the next pillar: content optimization.

Pillar #3: Content optimization

We follow the same content optimization principles for AEO and SEO.

In a nutshell, the content optimization elements look like this:

Although the process is the same, the playbook for AEO has changed based on how AI search engines surface the results.

Here are some patterns we noticed:

• LLMs prioritize early passages, which is why the way the top of your page looks makes all the difference.

• Chrome has a limit on the number of passages for embeddings, so you should keep intros short and straightforward.

• Long reads don’t generate infinite citations; thus focus matters more than length.

• Google’s AI follows the semantic HTML structure of a page and uses headings to understand meaning, which is why they shouldn’t feel ambiguous.

• Clear heading structure, such as the one asking a question and immediately giving the answer, makes content more skimmable for AI search engines.

• An AI search engine may retrieve a separate section, which is why it should be logically structured but not isolated from the rest of the content.

• The main priority is the reader, so aim for a natural flow with a clear structure rather than optimizing content solely for LLMs.

Based on these patterns, we developed a set of best practices that you can see below.

Do these practices work?

We applied them when writing this piece on the best AEO agencies:

• a clear structure

• straightforward intro and headings

• succinct and focused paragraphs

• lots of tables

• and zero fluff.

Here’s what we’ve got in return:

Of course, all of these best practices become largely decorative if crawl efficiency optimization gets ignored.

Let’s talk about it next.

Pillar #4: Crawl efficiency

Our crawl efficiency analysis is a 4-step process.

First, we check if AI search engines can access your pages, which you can easily do by going to your website’s robots.txt file and seeing if it restricts any user agents from crawling your website.

Here you can see that the website disallows AI agents from training on and crawling the website information, mitigating its chances to appear in AI search engine results.

The next step is creating (or checking the correctness of) the llms.txt file (or the AI Info page).

We should emphasize that adding this page to your website is not a guarantee for better AI bot crawlability, and Google says this page isn’t necessary for better AI bot crawlability.

Nevertheless, our experience working with clients shows otherwise, and a properly structured AI info page can make an incremental difference to your AI visibility.

So, how do you structure your AI info page?

Essentially, it boils down to what AI search engines should know about your company in a specific order:

Now, there are a few things to zoom in on here.

First and foremost, pay attention to brand disambiguation: ensure that an AI search engine doesn’t confuse your brand’s name with something else.

Take Notion, for example. It’s a well-known brand, but it’s also a word on its own, so their llms.txt file has to specify the nature of this brand.

Another important insight here is to be specific about your target audience.

Make sure you include both who you want to target and who is not the right match for your product.

Here’s an example of how we interpreted this section for Minuttia’s AI info page:

You also want to add sources that prove and verify your credibility, such as case studies or research.

For Minuttia’s AI info page, we added three prominent client cases and referred AI bots to our Case Studies website page for more information.

On a general note, you should aim for incremental improvements that would have a tangible impact on how you’re cited by AI search engines, and not chase overnight AI wins.

As for steps #3 and #4, we want to zero in on the best practices for the URL structure that improve crawl efficiency:

For the URL architecture, we follow two different approaches for blog post URLs and for cluster URLs.

In both approaches, we use three different types of keywords to structure URLs.

Here’s how these keyword types compare against each other.

The rules are simple for a blog post URL - you simply choose the keyword type based on the main query your post is targeting.

Here are some examples with 3 types of keywords:

• https://www.meilisearch.com/blog/rag-for-structured-data

• https://viral-loops.com/blog/how-to-build-a-community-around-a-product

• https://viral-loops.com/blog/coming-soon-page-examples

The picture looks a bit different for the topic cluster pages.

Essentially, you have the same keyword types, but they apply to the URL slugs corresponding to each subtopic found in the given topic cluster.

Here’s a visualization to make things clearer:

If you want a real-life example for the topic cluster, the Product Launch 101 piece by Viral Loops shows how to structure URLs for the phrase match.

Basically, you have:

• Core page: viral-loops.com/product-launch

• Cluster page: viral-loops.com/product-launch/marketing-plan

• Cluster page: viral-loops.com/product-launch/plan-templates

• Cluster page: viral-loops.com/product-launch/strategy, and so on.

What you want for both phrase and term/semantic match is to keep things simple. Avoid numbers, dates, or anything that would confuse the crawlers.

Now, let’s move on to content monitoring.

Pillar #5: Content monitoring

Let us return to the aforementioned finding by Ahrefs that AI search engines tend to cite fresher content more often. That’s a point to the importance of publishing velocity.

However, focusing too much on new content while ignoring outdated posts is a bit like renovating the kitchen while the basement floods.

Just look at the Peanut app’s website.

According to Ahrefs, Peanut’s blog contains 2,441 pages, which means they built them at scale, most likely posting over 100 blog posts a month.

This effort didn’t go unnoticed, and their organic traffic peaked at 2.5 million monthly clicks in 2023.

However, as their content inventory started to age (some content pieces haven’t been updated since 2022), organic traffic also took a hit.

Essentially, too much focus on net new content created a dead weight of older, sometimes irrelevant pieces, making regular content updates a real necessity.

However, a content update doesn’t necessarily mean a full rewrite.

At Minuttia, we developed a signal-based system that flags content due for an update. This system assigns value, which determines how much of that content needs to be revised:

• Light refresh: we change < 30% of content, making improvements to the specific sections, such as outdated statistics.

• Content update: > 30% of content is changed, involving a full rewrite for specific sections, e.g., when rankings have significantly declined, or the SERP landscape has changed.

• Complete rewrite: we rework the entire piece, mostly when it’s too outdated to salvage, or the search intent for the target keyword has changed.

What is important here is not to fully automate this process and always have a human in the loop.

The cost of a poor update is too high to delegate this task fully to, say, an AI assistant, so there should always be a team member who does quality assurance.

Let’s move on to the last pillar.

Pillar #6: Brand experience and perception

Now, if you go back to the intro to this section, you’ll see that the sixth pillar for SEO optimization mentions backlinks.

However, since the introduction of AI search, backlinks have become only one part of that pillar, yielding to the following for elements, brand experience being the most important.

Now, where do you start with improving brand experience?

You start with what you can control: owned perception of your brand.

Here are some examples of how we manage it: being listed in the directory, including a positioning page on our website, creating relevant blog posts, adding the AI Info page, and so on.

How does that translate to AI search visibility?

The more you invest in owned channels, the more likely AI search engines are to cite them in a branded prompt.

So, it makes sense to invest in multiple platforms for owned perception, but you want to do it in a way that matches your positioning, messaging, and business model.

What about the risk of sounding overpromotional?

It’s a good practice to develop or acquire digital assets to help with distribution - a newsletter, asset library, knowledge base - anything that corresponds to your brand’s positioning.

Now let’s talk about multi-channel perception. Why does it matter for AI search visibility?

The following data insights lay out the reasons perfectly.

• According to Ahrefs, YouTube is the most cited domain in AI Overviews: its presence has increased by 34% over the past six months.

• Kevin Indig said on his LinkedIn that more G2 reviews correlate with higher AI visibility.

• Ahrefs claims that brands getting the most web mentions earn up to 10x more mentions in AI Overviews.

Essentially, it all boils down to the number of brand mentions, and there are quite a few ways to get them without appearing too self-promotional.

But here’s the thing: you don’t always control the narrative when it comes to multi-channel perception. Nevertheless, when a negative brand mention comes along, it doesn’t mean you shouldn’t react.

Take the following example.

Asana, one of the leading work management platforms with 4.4 stars on G2, still gets negative reviews on Reddit. The one below complains about the unintuitive UI, but instead of leaving it to other Reddit users to solve, an Asana representative chimed in to assist with the problem.

This is one of the correct ways to handle such blows to your brand image. And overall, if you want to join a conversation about your company, we suggest doing it from a branded account, so your participation is transparent and trust isn’t undermined.

The same goes for discussing your brand in communities. It’s a bad practice to flood subreddits with promotional posts, like in the example below.

Instead, build or help moderate a relevant subreddit, where both your marketing and support teams can actively contribute without breaking trust.

Obviously, developing owned and multi-channel perception alongside other components of AI visibility is pointless unless you can tie all that to business outcomes.

Yet, AI search ROI and measuring AI performance, according to our survey, proves to be a big challenge for most respondents.

How to Measure AI Search Performance

If you look back a couple of chapters, you’ll see that in our survey, 40.6% of respondents reported measurement and attribution of AI search performance as their biggest challenge, followed by the lack of best practices and inability to tie AI search visibility to ROI.

It’s not entirely unexpected!

If you run the same prompt several times and across multiple AI agents, the results will vary, and even if your brand appeared once or twice, there is no guarantee it will be there the third time. How can you properly measure that?

There’s one crucial caveat to mention when it comes to AI performance metrics: measurement needs repeated sampling and context.

You need to see how things change and evolve over time, and it should be done often, given the volatility of AI search results.

When it comes to specific AI performance metrics, we think the following ones matter most:

• Referral traffic from AI platforms (e.g., ChatGPT, Perplexity, Claude)

• Brand mentions and visibility in AI search engines

• Share of voice in AI search engines

• Average position your brand appears in AI search engines against other brands

• Number of citations and citation share in AI search engines

• Conversions and revenue attributes to AI-referred traffic

• Sentiment

• Incrementality (h/t Kevin Indig)

• AI Crawler Activity

• AI Overview Impressions

Here’s a short description of each metric (how we perceive them):

Now we come to the point where those 40.6% of respondents feel the most pain: setting up the tracking tools.

Let’s make one thing clear: you’ll need to use several platforms and assign them to different metrics, such as GA4 for AI referral traffic or Google Search Console for AI Overview impressions.

Below, you can see how we triangulate across multiple data sources:

From these sources, GA4, Google Search Console, Bing Webmaster Tools, and server logs are already familiar to you. The question is, how do you set up an AI search tracking report?

At Minuttia, we follow this 8-step process.

As you can see, the first two steps involve the client because we need to figure out all the relevant topics and related prompts (go to Topic Breadth and Depth to read more about it). Then we break down these prompts into two categories:

• Topical prompts related to a use case, problem, or category in which a brand competes.

• Evaluation prompts that directly test AI’s perception of your brand (pricing, integration, customer profiles, etc.).

Most AI tracking tools focus on topical prompts alone, but the problem with that is they often give results for synthetic prompts, which don’t always reflect the way how people communicate in AI search engines.

That’s why you need a combination of topical and evaluation prompts to see the whole picture, but while topics are quite straightforward, it’s not always clear how to get evaluation prompts.

Allow us to let you in on how we do that.

The process runs like this:

And here’s the template for an evaluation prompt:

Evaluate [Brand] on the [platform/product/solution]’s ability to [specific evaluation criterion], [desired outcome or business impact].

You can type it into ChatGPT and then check whether the response matches the reality and if there are discrepancies to be fixed.

Note that here you’re measuring perception, not visibility. The primary goal of evaluation prompts is to see how AI search engines see you.

We also recommend checking the cited sources to understand if anything needs to be fixed or created to fill the perception gap, and do that for each AI search engine separately - different search engines mean different retrieval methods, sources, etc.

Which tools do we recommend to track the prompts and generate reports?

Honestly, there is no single tool that would do this job perfectly.

A go-to platform for any brand would be GA4, but what we found is that the data it shares often deviates from the real picture.

For example, in our experience, some AI-driven referral sessions are flagged as direct.

What we recommend you do instead is to create a separate AI Search channel in GA4 that groups ChatGPT, Perplexity, Claude, and any other AI search engines you want to track. But even then, the data can get misrepresented.

So, our closing point is that you shouldn’t take this data at face value and treat it as directional.

No decimal-point precision reflects reality but can indicate some trends, which ultimately make a difference for the bigger picture.

Maja here. The truth about AEO is that nobody has it fully figured out yet. The mechanics shift every month, the tools are still maturing, and what worked last quarter might not work next quarter. But the teams that start building muscle here now will have a real lead 12 months from now.

A few things worth doing this week:

• Pull up your robots.txt and check whether you’re accidentally blocking AI crawlers. Takes 5 minutes, and I’ve seen teams discover they had agents blocked they didn’t even know about.

• Pick two evaluation prompts (the ones that test how AI describes your brand and category) and run them across ChatGPT, Claude, and Perplexity. Compare what gets cited. That’s your perception gap, sitting right there in plain text.

Big thanks to George and the Minuttia team for putting this together. Drop your questions in the comments, or feel free to contact them directly.

If you want to dive deep in AEO, check out previous GTM articles on this topic:

• How to Win the New SEO Game: AI Search, GEO, and the Future of Visibility

• AEO: How to Make AI Recommend Your Product

See you next week.

Maja

✅ Need ready-to-use GTM assets and AI prompts? Get the 100-Step GTM Checklist with proven website templates, sales decks, landing pages, outbound sequences, LinkedIn post frameworks, email sequences, and 20+ workshops you can immediately run with your team.

📘 New to GTM? Learn fundamentals. Get my best-selling GTM Strategist book that helped 9,500+ companies to go to market with confidence - frameworks and online course included.

📈 My latest course: AI-Powered LinkedIn Growth System teaches the exact system I use to generate 7M+ impressions a year and 70% of my B2B pipeline.

🏅 Are you in charge of GTM and responsible for leading others? Grab the GTM Masterclass (6 hours of training, end-to-end GTM explained on examples, guided workshops) to get your team up and running in no time.

🤝 Want to work together? ⏩ Check out the options and let me know how we can join forces.

Supported by 1mind → AI Superhumans for revenue teams

Superhumans (1mind’s Agentic Conversational AI Agents) integrate seamlessly into existing workflows to qualify buyers, book meetings, deliver pitches, give demos, handle objections, uncover pain points, build value models, and onboard customers. They delight buyers, drive operational efficiency, and scale beyond human limitations focusing on increasing revenue, shortening sales cycles, and increasing ACV.

Founded by Amanda Kahlow, former founder and CEO of 6sense and a trailblazer in account-based marketing (ABM) who has now created AILG (AI Led Growth). 1mind has raised $40 million from Battery Ventures, Primary Ventures, Wing Venture Capital, Harmonic Growth Partners, Operator Collective and Success Venture Partners, and prominent executives from Monday, ZoomInfo and Databricks.

See the Superhumans in action

Dear GTM Strategist,

I used to think selling would be the last human frontier. The relationship work, the trust-building, the reading-the-room work - I assumed AI would help around the edges while humans stayed in the seat.

Then I met Nigel, 1mind’s newly launched Ride-Along Superhuman. My honest reaction: “I thought this would be possible by 2030. And yet, here we are.”

Here’s the thing most GTM leaders don’t talk about openly: the Solutions Engineer (bottleneck is the silent killer kof B2B deal velocity.

The median AE:SE ratio is 4:1. In many orgs, it’s 10:1. Which means most live sales calls have zero technical support in the room. The buyer asks a hard question, the rep stalls, and the deal goes cold. Qualified pipe doesn’t lose to competitors. It loses to friction.

Teams try to patch this with duct tape: AI listeners, Slack-based help channels. Fine for nudges, useless when the prospect asks something off-script.

The old answer is to hire more SEs. The smarter one is a two-layer SE org: AI handles the repeatable 60-70% of Q&A live on the call, humans handle the strategic 30-40% (where most SEs want to be anyway).

That’s the bet 1mind is making. They are aiming to scale the SE motion, not replace it. Boston Consulting Group predicted this in their August 2024 The Future of Sales with AI report. 1mind is one of the first to deliver on it.

Nigel reads the room like a seasoned pro, stays focused on what he was trained for, and never bluffs. Which is more than I can say for most human-led demos I’ve sat through.

In this edition, I invited Jonathan Kvarfordt to join forces again. Jonathan has a great eye for winners and is one of the sharpest minds in our space.

Here’s what you’ll learn in this edition:

• Why the standard SaaS buyer journey is structurally broken, not just slow

• The OAR Matrix: how to decide whether to Optimize, Amplify, or Reinvent with AI

• The four lanes every GTM team will pick from in the next 18 months

• Three concrete moves every GTM leader should make this quarter

• What “AI-first GTM” actually looks like in practice, with proof points

Jonathan, the floor is yours 🎤

The buyer journey was built for 2018. Your buyers live in 2026.

Walk through your own funnel as a buyer. The pattern is painfully consistent.

A prospect researches their problem on Perplexity or ChatGPT. They have recommendations from AI and then they land on your site, like what they see, and fill out the inbound form. A 22-year-old BDR who started six weeks ago calls them and asks the same qualifying questions Perplexity already answered.

They get scheduled into discovery with an AE, who asks most of the same questions again.

Meanwhile the buyer is pinging peers and former users to self-educate, because nobody on your side has actually answered their question yet. Eventually, a demo happens. Then come the follow-ups, the stakeholder alignment calls, and - if everyone is still awake - a close.

Sound familiar? It should. This is the standard SaaS playbook, and it has a clear lineage. Predictable Revenue in 2012 gave us the SDR/AE split. SalesLoft, Outreach, and 6sense around 2014 gave us sequenced cadences, lead scoring, and intent data. 2020 forced the entire industry online. 2022 added the PLG layer.

By 2018, the model had settled: deterministic workflows, MQL-to-SQL handoffs, BANT-or-MEDDIC qualification gates, seat-based pricing, and 80-90% gross margins that made it all economically viable. That model was beautifully tuned for its world.

And then came AI-led growth (AiLG) and changed everything. Buyers self-educate before they ever talk to us, AI has compressed research time from weeks to minutes, gross margins on AI-native products have collapsed to 50-60%, and the buyer increasingly knows more about your category than your rep does.

We are running a playbook designed for a world that no longer exists. It is yielding diminishing returns and is expensive to run.

How to fix it?

Humans are only capable of so much and 75% of the drag is internal friction anyway

Here is the part of this conversation almost nobody is naming.

When my team at Momentum ran the analysis, 75% of the friction slowing down sales cycles came from inside the company, not from the buyer. It wasn’t RFP processes, security review, or buyer indecision dragging deals out. It was internal handoffs, approval queues, and CRM hygiene. SDRs reminding AEs, AEs chasing SEs, SEs waiting on implementation.

The deal moves slowly because we move slowly.

And the cost model is breaking right alongside it. Depending on stage and industry:

• 10-50% of CAC goes to sales and marketing

• 20-50% of revenue is spent on sales alone

• 50-80% of cash burn is driven by headcount

So we’re running an expensive, slow model in a market that rewards the opposite.

Most GTM teams aren’t built for the AI world. Handing your 2018 playbook a ChatGPT license is not transformation. You do not transform from writing emails faster.

Most “AI in GTM” conversations stop at the buyer-facing surface - chatbots, AI demos, automated outreach. That’s treating the symptom, not the system. You cannot fix what’s broken by trying harder on top of it.

Let me be very clear here. The old world would say, “hire more humans” but countless thought leaders, analysts, and companies in the space have already identified that growth at all costs does not work.

Here is the other side of the coin: even if you could have all the funding in the world, humans have capacity constraints. Knowledge limits, only work at certain times, talent quality variability, scheduling nightmares, etc. These constraints cause, as we have talked about, which is the horrible buying experience. So what if we looked at the issue from first principle thinking? What if we ask the question, what is best for the buyer and customer experience?

After working with thousands of individuals, hundreds of companies, and working on AI strategy and implementation since before ChatGPT was released, I can tell you, you can only optimize a human so much because of the constraints, at some point, you hit a ceiling. When I was at Momentum, we released the book called Ignite, where we interviewed 30+ experts in the space about this and everyone came to the same conclusion, which is to think in systems and that means balance between humans and AI, not only AI and not just only humans - it’s a harmony between the two. Architecting the intelligence of both humans and AI, not one replacing the other.

At 1mind, every use case we solve for is something that humans do not do right now anyway. For example, how many companies do you know of that have an expert human on their website (not an SDR), ready to jump on a Zoom call at any time and able to answer deep technical questions 24/7?

The answer is… no one. Why? Because no one can afford it.

How many companies have 1:1 coverage of AE to SE? Not many… why? You got it, no one can afford it, let alone find enough talent to fill the gaps, even if they could.

How many companies have a product expert ready to answer deep technical questions right after someone buys a product? Again, not many.. Most give you a few hours of onboarding or some sort of onboarding doc and send you on your way…

How many Customer Success teams have deep technical support anytime they need it?

If anything, every customer success leader and team I have been on has been begging for more help, not less.

But what if we ask the question, what if we gave a genuinely delightful experience? What if we gave our BEST to our customers from the start? What if we used technology as a way to make the customer experience better? Is that not a good thing?

This is me as a human actually typing a phrase that AI usually does, but this is NOT about replacing humans, it is about doing things that humans are not doing right now in the first place.

What to do instead?

The OAR Matrix: put AI where it actually belongs in your org

Most GTM teams I work with want to jump straight to “AI transformation.” The smarter move is to know which level you are actually playing at.

At GTM AI Academy, which has helped over 10,000 leaders on their AI Adoption Journey, I teach a simple framework called the OAR Matrix to help them figure out where they are and where they want to head next.

1. Optimize. Same process, made faster with AI. Realistic gain: 10-30% time savings. For a large org, that is a serious ROI number. Start here if your team has low AI maturity or low appetite for disruption.

2. Amplify. Real structural changes. More revenue, real cost savings. Realistic gain: 20-40%. You are not blowing up the function, but you are seriously rewiring it.

3. Reinvent. Burn the playbook. Rebuild from first principles. Realistic gain: 100x on the right function.

Most organizations are optimizing where they should be reinventing. You may be asking, but will humans actually talk to an AI and get value? Will businesses actually use this?

To make that concrete, look at HubSpot - one of the most sophisticated demand engines in B2B.

HubSpot could have gotten stuck on the Optimize path: better sequences, ChatGPT licenses for the SDR team, faster routing. Instead, they put an AI Superhuman named Fiona at the front of their funnel and let her handle the buyer conversation end-to-end.

The results speak for themselves:

• 88% of buyers who landed on the page engaged with Fiona

• 32% had a meaningful conversation - averaging nearly 8 minutes, with 90% surfacing real pain and 55% getting into pricing

• 78% increase in free trial conversions

• 25% increase in influenced pipeline

But the number that should stop every RevOps leader cold is this:

HubSpot estimated it would have taken 83 SDRs and 19 sales engineers to deliver the same volume and depth of buyer conversations.

Remember, this putting AI in a place that HubSpot could not hire for in the first place, so this comparison is not saying they replaced this many people, they identified how many people it would have taken them to hire to be comparable.

That is the gap between Optimize and Reinvent. Optimize gives you 20% more from your current 102-person team. Reinvent reimagines the entire process where AI gets put into the mix for a better buyer experience and allows humans to do more of what you need your humans to do, like build better relationships, stakeholder management, and connect with customers the way that only humans can.

Reinvention always feels like a slingshot pulling backward. It takes time to reinvent and build from the foundation up. So you feel like a rock in a sling shot and while everyone is on Linkedin posting about the latest agent that made them $2 million in 2 days that they built over the weekend, you are feeling like you are falling behind. But trust me, when you do this right, you accelerate faster than ever. You don’t catch up to where you were, you launch past it.

In 2027, buyers will choose how they want to buy from you

Up to this point, I have been talking about what AI does inside your GTM team. Here is the harder question: what AI does to the buyer’s experience of you.

Within the next 18 months, every company will land in one of four lanes. There is no neutral position, and the lane you pick becomes part of your brand whether you intend it to or not. Because the BUYER is the one who will control how they want to interact with you, because now they have options. Which means we as companies need to think about giving them a “choose your own journey” type of experience.

For example:

1. Human only. The buyer demands human only experiences. For whatever reason some humans will reject AI and demand a human only experience. No one knows the size of this group, but it will be a reality at least for the coming future. If this is the case, those human buyers will expect the best of the best. Mediocrity will not be tolerated because just like how Amazon released 2 days shipping and you expect every company you work with to have 2 day shipping, buyers will have the same expectations when it comes to overall experience. The Tide lifts all boats and that tide is a better buying experience is a minimum requirement.

2. Human, with AI assist. The buyer goes through a similar experience to what most have right now where the seller stays in the seat; AI handles the back-office work. This is Optimize on the OAR Matrix. Most GTM teams I work with are aiming here - realistic, achievable, a solid place to operate from. AI is here and there in the process, but it focuses on human-to-human interaction.

3. Human to AI. This is where things start to shift into AI Led Growth. Gartner surveyed 646 B2B buyers and identified that 67% PREFER a rep-free buying experience. If someone is willing to work with your company but is tired of the old-school buying process that we already walked through, why not give them the option of doing more with AI? HubSpot is running this. Seismic is running this. ZoomInfo is running this. The bet is that buyers want optionality, and the data so far says they do not only in surveys, but in real life.

4. AI to AI. This is the one most people are not ready for. Within 12 to 18 months, a meaningful share of your inbound traffic will be AI agents - OpenClaw, Hermes, or self built Claude Code or Cowork or ChatGPT AI Agents all acting on behalf of a human buyer, doing the research, comparing vendors, and in some cases making the purchase recommendation. Is your website, your pricing page, your demo flow designed to be evaluated by an AI agent in 90 seconds? The future will be where the human buyer has a fleet of agents not only doing research, but has a wallet with a budget that they can pay for solutions. Are you ready?

The strategic question is not “will AI replace my salespeople.” That is the wrong question, and it makes leaders defensive. The right question is: what buying experience does our brand want to be known for? Pick your best fit. Design for it.

The principle underneath it all: design around the buyer, not the seller

The single biggest mindset shift required to operate in any of the four lanes above is this: stop designing your GTM motion around what the seller does, and start designing it around what the buyer needs at each step.

Once you do, you discover that a lot of seller activity exists only because the seller’s workflow demanded it - the BDR call, the AE re-discovery, the qualification gate. AI handles the parts the buyer never needed a human for. Humans show up where buyers actually want them: judgment, negotiation, edge cases, relationship.

That is the actual reorganization. Not your org chart. Your workflows.

Three things to do this week

You do not need to rebuild your GTM org by Friday, but you do need to start. Three moves:

1. Walk your own funnel. Pick a friend in your industry and have them go through your buyer journey end to end - inbound form, BDR call, AE call, demo, follow-up. Measure how much time it takes to get to the end goal. Have them rate the experience 1 to 10 and answer one question: would they recommend it to a friend? Most GTM leaders have never done this exercise on their own funnel. Most are stunned by what they find.

2. Map the internal handoffs, not just the buyer journey. Where does the deal slow down inside your walls? SDR to AE? AE to SE? Approval queues? Quote desk? Pricing exceptions? Map every handoff and time-stamp each one for your last ten closed deals. The pattern will be obvious within an hour, and it will tell you exactly where AI belongs first especially when you map these frictions to your buyers journey. Optimize in areas where these frictions overlap.

3. Pick one workflow to redesign around the buyer. Don’t try to fix everything. Pick the single most expensive workflow in your funnel - your AE discovery process, your trial-to-paid motion, your renewal flow - and ask which steps exist because the buyer needs them, and which exist because your process demanded them. In that area, give the buyer and customer the option of AI. You will be surprised how many will not only use the AI, but prefer it.

A year from now, the GTM orgs that win will look smaller, faster, and weirder - fewer humans, doing the parts only humans can do, with AI doing the rest. The org chart won’t look like 2018. Neither will the revenue per head. This will cause disruption of the old world

And one more thing

You can read about AI Superhumans all day. The only way to understand what one actually feels like in production is to have a conversation with one. Mindy at 1mind, Fiona at HubSpot, and others are running live - go talk to them for five minutes.

It will tell you more about where GTM is headed than another article will.

Explore the Customer Superhumans →

The question worth sitting with

What buying experience does your brand want to be known for in 2027 and beyond?

Because if you don’t decide, your 2018 playbook is deciding for you.

✅ Need ready-to-use GTM assets and AI prompts? Get the 100-Step GTM Checklist with proven website templates, sales decks, landing pages, outbound sequences, LinkedIn post frameworks, email sequences, and 20+ workshops you can immediately run with your team.

📘 New to GTM? Learn fundamentals. Get my best-selling GTM Strategist book that helped 9,500+ companies to go to market with confidence - frameworks and online course included.

📈 My latest course: AI-Powered LinkedIn Growth System teaches the exact system I use to generate 7M+ impressions a year and 70% of my B2B pipeline.

🏅 Are you in charge of GTM and responsible for leading others? Grab the GTM Masterclass (6 hours of training, end-to-end GTM explained on examples, guided workshops) to get your team up and running in no time.

🤝 Want to work together? ⏩ Check out the options and let me know how we can join forces.

This newsletter is sponsored by Miro.

Miro’s MCP server connects your boards to the AI tools you already use: Claude, Cursor, Lovable, Replit, ChatGPT, and more. This turns your collaborative workspace into a live context that AI agents can read, research, act on, and build from.

Imagine sending your prototype, which you agreed on with your team in Miro, directly to Claude Code to build it in minutes — that’s a reality with Miro’s MCP server (and much more).

I’ll be at Canvas 26 London 🇬🇧 on June 2 to be the first to know what Miro ships next. I will also host a workshop there at The Studio on how I use Miro. If you’re going, let’s totally shake hands. I’d love to compare notes on how you’re putting collaborative AI to work in your GTM motion. The San Francisco keynote will also be streamed on Miro’s YouTube account, which is free to watch.

Save your spot

Dear GTM Strategist,

It is easy to spend your time looking at people further ahead. But every milestone deserves a pause. I just crossed 33K subscribers on Substack and 83K on LinkedIn, with 1 million monthly impressions, in two and a half years of consistent content creation.

None of it was an accident, and none of it was linear.

It started in November 2023 when I launched my book, The Go-to-Market Strategist. I had spent months prepping for the launch, and I had no idea what would happen. Luckily, it became a bestseller. That was the aspiration. But then: now what?

So I started a newsletter on Substack, mostly because the platform had embedded growth features and a few content creators I admire were using it. 🤷 No sophisticated agenda there. I also kept up the habit of posting three to five times a week on LinkedIn, since it had worked well for the book launch sign-up list. Why stop?

I want to be honest about how this actually happened, because most “how I grew my audience” content makes it look linear and inevitable. It was not. There was a 6-month period in mid-2025 where I genuinely lost the plot. I will get to that.

In this article, I will walk you through:

• The 7 growth levers behind GTM Strategist, in the order I built them

• The mid-2025 identity crisis that broke the pattern and forced a rebuild

• A simple 3-layer framework for thinking about your own lever stack

• What I am betting on next in AI-native GTM

So let’s get started. We have a lot to cover.

7 levers in detail with useful how-to content

Before I walk through each lever, here is the shape of the journey. Every lever, once added, kept compounding. None of them replaced what came before.

What is interesting in this curve is what happened between February and September 2025. I was doing all the right things, and the growth flattened anyway. That was the period that broke me a little, and it was also the period that taught me the most.

Let me walk you through it.

Lever 1: Voice and Consistency (The Longest Reps)

The first 6 months after the book launch, I was just learning by doing.

Looking back, I’m not sure what I was doing exactly. But I was doing the work. One of the great things about GTM is learning to play to your strengths and compensate for your weaknesses. My strength is being extremely hardworking and disciplined. Copywriting was never my passion.

I am a long-form girl. I don’t spend my days thinking about hooks, and I have a philosophical barrier to copying what direct-response copywriters do - it feels repulsive to me as a person. I acknowledge the discipline is highly effective for conversion-centric copy, and I will happily pay someone else to do it, but I don’t have it in me.

Luckily, social media and newsletter content are a different beast. You need something proprietary to say - and in the world of AI-generated everything, that matters more than ever.

My first instinct was to promote sections of the book in the newsletter. That worked for a while, but I quickly realized that not all the interesting GTM knowledge in the world lives between two book covers. So I started exploring topics that came up with my consulting clients and inside our community.

What I underestimated back then was how long it takes to find a voice that actually sounds like you. Not a “brand voice document” voice. A voice where people read three sentences and know it is you.

Today, you can shortcut a lot of this by embedding call recordings into your content AI engine. Back then, AI-assisted writing was barely a thing - ChatGPT had just properly launched - and I tried 6 different copywriters before I gave up. None of them could mimic my voice well enough for me to be happy with the output. So I sucked it up, developed just enough copywriting muscle to get by, and committed to writing everything myself, every single day for six months.

The reps are the moat. There is no version of this where you skip the early uncomfortable months.

🔗 Dive deeper: I documented the full first year of building GTM Strategist here - a useful read if you are prepping for a launch or building your own content engine backed by digital products.

Lever 2: Design as a Moat

When copywriters and AIs could not replicate my voice, I made a strategic call: I would write everything myself and double down on design instead.

This turned out to be one of the best decisions I made. Back then, most content creators were either DIYing something in Canva or paying stunning amounts ($200+) for a single infographic that honestly looked the same to me every time. I get it: don’t fix what’s not broken, but also innovate because people are getting saturated with formats.

At the same time, most newsletters in our space looked the same - blocks of text with the occasional stock photo or screenshot. Investing in custom illustrations, frameworks, and visual systems made my content instantly recognizable in a LinkedIn feed.

Design became my first real differentiator, and it still is.

These days, the GTM Strategist design system is pretty sophisticated. We have fundamental brand assets and a brand book, and we are capturing great trends all the time. As we experimented with new formats, topics, and visual aesthetics, it became a little design lab.

🎁 Gift number 1: Based on the analysis of my best-performing LinkedIn posts in the last 90 days, here is a cheat sheet for designs that work well for me. The analysis was done with Miro’s MCP in Claude. Get it here.

The Big Lesson: When everyone else is racing toward the same AI-generated aesthetic (currently, orange Claude-generated images, even though ChatGPT became amazing at infographics), the things that are harder to copy (aesthetic illustration, complex graphics, atypical layouts) become more valuable. While I know plenty of people who are doing great work with AI-generated designs, I am continuing to invest in human-led design and branding.

Lever 3: Collaborations and Cross-Recommendations

You can go far on your own. You can go further with the right people by your side. Or, more simply: go where your audience already is.

I was such an underdog when I started. I am originally from Europe, English is my second language, and we have a slightly unconventional take on GTM here. Some would call it scrappy. I call it extremely pragmatic - we get the job done. The first year, I was not really taken seriously by the US-centric GTM crowd. The “big guys” mostly looked past or down at me.

But I was eager to show them I was not a force to be taken lightly. Not only can I outwork most people in the space, I genuinely believe I have good advice beyond theoretical fluff, built on solid fundamentals. Expanding the community and evolving from scrappy, bootstrapped startup roots toward an enterprise-relevant, GTM-first approach became a really important mission.

I am forever grateful to Kyle Poyar, Aakash Gupta, Aatir Abdul Rauf, David Pereira, and Pawel Huryn - among the first “big guys” who took me seriously. They invited me to co-create content, gave me real advice on how to grow, and helped me develop a more sustainable content flow. Not one that was rooted only in my product portfolio, but one that was starting to become a recognizable voice in the space.

People ask me all the time how to find this kind of person. My rule: I personally have to love their content and admire them as professionals.

How to find your people:

1. Ask your existing audience who else they follow and learn from in the space.

2. Use desktop research or AI research to map adjacent creators and operators.

Identifying them is not rocket science. Deserving their attention is the harder part. Sometimes it takes me months, even years, to build these relationships. Dozens of multi-channel follow-ups. The slow grind of becoming useful before becoming visible.

But once you find a real synergy, those partnerships become invaluable. They are the partnerships that shape ecosystems.

If you are building a brand right now: do not optimize for the algorithm before you have optimized for peer trust. The algorithm changes every quarter. Peer relationships compound for years.

I am still in touch with nearly everyone who supported me early on. We learn and grow together. Some sweet collabs coming up in the next couple of months. Stay tuned. 🫶

Lever 4: Free Assets That Feel Like Paid Products

In 2024 and early 2025, the “comment to get X” era on LinkedIn was in full bloom. Most of those posts were thinly veiled email harvesters. The asset itself was rarely worth the comment.

I took the opposite approach, almost by accident. The principle, in retrospect, is product-led growth applied to content: give away something for free that people would happily pay for.

This brought real, relevant attention to my profile. Not vanity engagement. The kind of attention that turned into clients, partners, and long-term subscribers.

There was a temptation around this time to go mainstream. The content creators I admired when I started out were shifting toward career, leadership, and personal-growth content. The reach of that kind of content was undeniable. But that was never my passion. I am obsessed with GTM.

Fast forward to 2026. “Comment to get X” is making a comeback because AI has made deep research available to an average student. Anyone can ship a free guide in a weekend. Which means the bar is back to where it should have been all along: the asset has to actually be good.

When people want something, it usually pays off to go the extra mile and create a special asset for them. Be selective. Not all followers are equally valuable to your business. Plenty of people would love anything for free.

What matters more than ever is ICP fit. If you have to go the extra mile to make something deeper, more meaningful, more researched - do it. If you go with mainstream blockbusters like “Comment to Get My Productivity OS,” and you work on neural computing consulting, it will not help grow your business.

The point is to find the right audience, not to boost vanity metrics.

🎁 Gift number 2: A little stash of my past “comment to get X” assets, in case you missed them on LinkedIn.

🔗 Dive deeper: My 5-step LinkedIn viral post framework - “comment to get X” does not work the way it used to, but when you nail the topic, it is still magic.

Lever 5: You Are the Niche (And Why I Almost Walked Away)

This lever overlaps with Lever 3, but it deserves its own section because it almost made me quit the whole thing.

The temptation to broaden the range of topics was real. You can easily get thousands of likes on LinkedIn by oversharing your personal life, riding generic leadership trends 🥱, or quoting Steve Jobs for the millionth time.

Big content creators call them “audience growers”.

I call them “if you see these three times, hit unfollow”.

What kept me grounded was a simple realization: I am obsessed with GTM. The day my content stops being about GTM is the day I stop showing up with conviction, and readers can feel that.

This cost me time. It probably cost me follower count. But it built the only thing that actually matters long-term: a reputation as a thought leader in the space. You cannot build that by accident, and you cannot build it while chasing every adjacent trend.

And then, for a couple of months in 2025, I lost touch with that.

When everything stopped working 🤯

Now we get to the part most growth stories skip.

Mid-2025 was very difficult for me, not only as a content creator but as a person. I was struggling to find my voice because mainstream AI had arrived with these grand promises of agents doing everything for us. I am a big believer in AI technologies, but I was not (and still am not) convinced we are at that point yet.

Around the same time, the LinkedIn algorithm shifted. Posts that used to hit 20K impressions consistently started struggling to reach 10K. The same effort. The same topics. The same care. Half the reach.

I was tired. I was frustrated. The constant algorithm games took a real toll on me. It became a negative spiral.

The Substack growth flattened too. You can see it clearly in the chart above. February to September 2025 is a near-plateau, even though I was publishing consistently.

But there is no nice way to say this: I lost a lot of my enthusiasm. I did not know what I was doing wrong. The honest answer, in retrospect, is that I was not doing much wrong. I just could not figure out fast enough how to adjust. The ground had shifted underneath me, and the system I had built was no longer the right system for what was about to happen next.

I needed a break.

My husband and business partner Anže, and I packed our bags and spent a month in Indonesia. It was fun. We came back with a clearer head, but also a little empty-handed. I was not sure what came next.

The answer came from two directions at once.

Lever 6: Curated Brand Partnerships (That Add Value)

Before I go further, I want to be transparent about something.

Brand partnerships are how I keep this content free and accessible to everyone who needs these insights. No paywall, no premium tier, no “subscribe to unlock.” Just one direct trade: I work with a small, curated group of partners, and in exchange, the work stays open for the entire community. Huge thanks to all our partners 🫶

That is the deal. And because that deal matters to me, I take partnerships very seriously.

After Indonesia, I went back to first principles: give people what they want.

I talked to community members. I reflected on what content I actually enjoy consuming. The conclusion was simple. Everything has to be useful, hands-on, and well-researched. Especially now that creating content has become so easy, having something smart and useful to say is the only sustainable way to remain relevant.

But the bigger shift was how I started thinking about brand partnerships.

I stopped seeing partners as sponsors who get a shout-out in exchange for a banner. I started seeing them as part of a GTM Strategist ecosystem - a curated tool stack that produces real ROI and sustainable work for the community.

Sourcing from our consulting work, I invited a small group of tools (up to 6 a quarter) we already use and trust to build long-term partnerships with us. Not for one-off campaigns, but for ongoing collaboration. In return, we get a steady pulse on what is new in their ecosystem, and our community gets something genuinely useful: tactical, practical content built around tools that actually work together.

Some of our best-performing content in the last 90 days has been partnership content - the “how to use X” deep dives, the workflows, the real implementations. Every piece is researched. Some experiments will work, some will not, and that is fine. The decisive call was this: stop shying away from naming the tools, and be ruthlessly honest about which ones we recommend.

The results have been strong on the partnership side, too. UTM links never capture the real value, even though we are consistently in the top 3 to 5 content creators for almost every brand we collaborate with. That is not me bragging. That is me saying we hold ourselves to a measurable standard, and we walk away from partnerships that do not clear it.

I get pitched at least three times a day for brand collaborations. Most of them, I redirect to other creators who would be a better fit. It does pain me to say no, especially to smaller tools and to people I personally admire. But sustainability and trust come first.

It took me years to build trust with this audience. It would take one or two bad partnerships to lose it.

So I am almost annoyingly selective. I try every tool myself. I will not feature anything I would not personally recommend to a client or use in our business for years.

Lever 7: Riding the Claude Code Wave (Early)

The second answer came at the start of 2026. And it changed everything.

Claude Code went mainstream. I was lucky to get into the ecosystem early - and I made a deliberate choice to share what we were learning openly as we figured it out, instead of waiting for a polished playbook.

The result: 3 to 5x my usual growth benchmarks in 90 days.

Some of my best-performing assets in that window have been built around Claude Code, context engineering, and what I am now calling the GTM brain - a proprietary AI-native GTM system that compounds rather than degrades with every model update.

🔗 Dive deeper:

• The GTM repository for Claude Code

• The GTM guide to AI context engineering

• Build your GTM brain (a demo with Attio)

That last part - “compounds rather than degrades with every model update” - matters more than it sounds.

For content writing specifically, we moved everything into the Anthropic ecosystem and built a system that consistently produces great results without getting worse every time a new model ships. I still firmly believe in human editing, but for the first time, I am genuinely happy with how we are innovating in content creation.

Did we burn too many nights and weekends in Claude Code? Yes.

Was it worth it? Yes.

The other unlock was the rest of the stack. Claude Code for mockups. ChatGPT for infographics. Wispr Flow for voice-to-text. I can speak faster than I can type, and Wispr Flow lets me move at the speed of thought. Not affiliated - just genuinely obsessed.

The pattern here is the same as Lever 3 with collaborations: be early to the platforms that are about to matter for your audience, and share what you learn openly as you figure it out.

Major Unlocks on this journey: The Lever Stack, in 3 Layers, One Moat 🏰

Looking back, I can group the 7 levers into 3 layers.

This is the framework I would hand my November 2023 self if I had a time machine. I do not, so I am handing it to you instead - in the hope that your demand generation journey will be a little smoother than mine.

The foundation layer - voice and design - is the part you cannot skip. Everything else rests on it. (Great candidates for your .md files, by the way. 😉)

The distribution layer - collaborations, free assets, niche discipline - is how your work reaches the right people.

The compounding layer - curated partnerships, platform bets - is where the brand starts working for you instead of the other way around.

Most people try to start at the top of the stack. They want partnerships before they have a voice. They want platform bets before they have an audience. It does not work that way. Each layer rests on the one below it.

Together, they compound into a distribution moat 🏰.

This is the holy grail of AI-native GTM. Products have never been easier to copy and build, so the speed of traction is the name of the game. And almost no one builds that moat overnight. It takes commitment, iteration, and the kind of dedication that does not photograph well.

That is the annoying part. It is also the part that makes the moat real - and the part AI can actually help you with, if you let it.

In the meantime, take care, and good luck building your own stack.

Catch you at the next milestone (50K newsletter x 100K LinkedIn, hopefully) 🤞

Maja

✅ Need ready-to-use GTM assets and AI prompts? Get the 100-Step GTM Checklist with proven website templates, sales decks, landing pages, outbound sequences, LinkedIn post frameworks, email sequences, and 20+ workshops you can immediately run with your team.

📘 New to GTM? Learn fundamentals. Get my best-selling GTM Strategist book that helped 9,500+ companies to go to market with confidence - frameworks and online course included.

📈 My latest course: AI-Powered LinkedIn Growth System teaches the exact system I use to generate 7M+ impressions a year and 70% of my B2B pipeline.

🏅 Are you in charge of GTM and responsible for leading others? Grab the GTM Masterclass (6 hours of training, end-to-end GTM explained on examples, guided workshops) to get your team up and running in no time.

🤝 Want to work together? ⏩ Check out the options and let me know how we can join forces.

This newsletter is supported by Attio, the AI CRM.

I helped build something every founder should have in their toolkit.

GTM Atlas is a free resource for early-stage teams building from scratch, and I wrote an entry on building a GTM brain that gets smarter over time, using context engineering to make AI permanently smarter about your ICP.

With Atlas, you’ll get a point of view on your growth strategy, frameworks you’ll return to, and the creative latitude to build the right motion for your business, not someone else’s.

Wherever you are, Atlas was made for you.

Mapped by operators. Curated by Attio.

Read my entry now

Dear GTM Strategist,

Lately, we have been so deep in Claude that we might have slightly neglected the “strategy part” of the go-to-market strategy in our conversations.

Reality check:

If we spend all our time copying each other’s repos, workflows, and skills- who is really driving our GTM? AIs?

Time to take a mini break from terminals and .md files and shift our focus from how to do X (tactics) to what to do (strategy) instead.

As the world shifts into the third wave of AI and AI-native companies disrupt our imagination of what can be done in record time with a handful of people (and agents), it is harder than ever to stand out, stay relevant, and build any moat.

To sharpen our strategic muscle, I curated 3 thought-provoking ideas from GTM Atlas:

1. Your product is the pitch by Elena Verna

2. Your website is middle of the funnel now by Emily Kramer

3. An AI’s job is to remove wasted motion by Kyle Norton

Some snippets made me feel really uncomfortable (no spoilers), which is the whole point of learning something new.

My sole intention for this piece is to slightly provoke your mind, prompt you to think differently, and arm you with relevant insights that could help you build a more effective GTM system.

Let’s see if I can make it happen.

Here are the big ideas I hand-selected for you:

1. Your Product Is the Pitch

Author: Elena Verna, Head of Growth at Lovable

The companies winning right now are giving the most away upfront - product, access, experience - before asking for anything in return. Decision makers (CISOs, CTOs, VPs) are doing their own evaluation. If you’re not letting them touch the product, you’re already out.

But here’s the part I keep coming back to.

“Stop trying to minimize the freemium. Treat it as the line item that replaces your Google and Meta spend.”

Lovable gives away millions of dollars of software every year. Hackathons, schools, book clubs - no requirements, no forms. Elena says it’s their best acquisition channel by a significant margin compared to any paid spend they do.

This is a complete reframe of how most founders I talk to think about freemium. They treat free usage as a leak to plug. Elena treats it as a marketing budget. Same dollar amount, totally different mental model.

A few other ideas from her chapter that I dog-eared:

• Satellite apps are replacing gated PDFs (and that is great for AEO 👌). Nobody wants to trade their email for a whitepaper anymore. Elena is seeing a rise in interactive, clickable mini-apps - real experiences with a back-end, where prospects enter their own information and get something useful back. It puts marketing in an engineering position, but the engagement is incomparable to flat content.

• Build in public is now an acquisition channel. When software becomes commoditized, what differentiates you is the team building it. Lovable’s first few million in ARR came largely from their CEO being publicly transparent about what was working and what wasn’t. Founder social brand is no longer a vanity project.

• AI agents are part of your ICP now. This one stopped me cold. Elena’s hypothesis is that B2B software is evolving to be consumed by machines, not humans. Humans will own the agent interface. Agents will perform the tasks. Your future buyer profile includes the machines doing the work.

If you’ve been treating PLG as something you’ll get to “after we lock in sales-led motion and start generating predictable revenue,” this chapter might make you feel uncomfortable. Which is the point. In the AI-first GTM world, speed is of utmost importance.

2. Your website is middle of the funnel now

Author: Emily Kramer, founder at MKT1

I’d never say that marketers are lazy. They are usually one of the hardest-working people I see in the organizations we work with. But all the busyness could be a trap! Many of them are so busy thinking about the next campaign, event, and webinar that they fail to see that the nature of their work has changed tremendously. The big question here is - are we busy doing things that still work?

Emily Kramer placed a fair warning: RAM - Random Acts of Marketing. It’s what happens when teams have access to more tools, more channels, and more AI-assisted ways to create content than ever before, and they respond by just doing more of everything. Copying what other teams do. Saying yes to every founder request. Never stopping to ask which box is actually worth checking. Emily’s argument is that the teams pulling ahead aren’t doing more. They’re doing less, better. How?

Her solution is the gen marketer: a generalist who can orchestrate across all the marketing sub-functions, run “fuel + engine” campaigns (where fuel is the content and engine is how you distribute it), think in big bets that differentiate, create leverage by tapping into ecosystems, and is genuinely fluent in generative AI workflows. The big question here is not how to do more, but how to extract more value from what we are already doing.

Of course, you want an example: The role of the website in the AI-first GTM world.

LLMs scour the internet and find your pricing, whether you put it on the page or not. Traffic is fundamentally lower because people are getting answers from chatbots before they ever click through. So when someone lands on your site as an arguably higher-intent lead, the stakes for converting them are higher than ever. Emily offers some practical tips: Stop gating. Put pricing in a structured format. Put meeting booking in the demo flow. The website is no longer where you figure out who your customer is - you should already know.

And this change has two massive implications for GTM down the line:

• Accounts before contacts. Emily is direct about this: we’re so obsessed with leads that we don’t think nearly enough about the accounts we want. It’s a multi-person sale. Stop net fishing. Do account research, niche down, and systematically pursue the ones that fit. Stop thinking about lead capture and qualification - start thinking about which accounts are even aware of you and what stage they’re at.

• Kill MQL and SQL. When you’re going after whole accounts with multiple stakeholders, MQL/SQL labels create territorialism between marketing and sales that’s completely unnecessary. The buying journey isn’t linear. These are milestones, not handoffs. The fight over “marketing didn’t deliver enough” vs. “sales isn’t converting enough” is a distraction from the actual job: getting qualified, hand-raising buyers and going full force on them together.

Maybe sales and marketing will finally become best buddies in this new world … or maybe agents will take over the customer journey - who knows.

3. AI’s job is to remove wasted motion

Author: Kyle Norton, CRO at Owner.com

And for those two who are 🥱 - too much growth and marketing - Kyle comes to the rescue with some thought-provoking numbers 🍽️ next:

“Our ACV is only $8 to $10k, so how do you make cold outbound work in that world? By obsessing over the quality of the data, and stripping every wasted motion out of your rep’s day.

Outbound won’t work if your BDRs are spending 70% of their day researching leads and calling contacts who aren’t the real decision makers. You can maybe book a meeting every two days. You press reps so hard on a dial number of a few hundred a day that they can’t find enough high-quality leads to hit it. If you tiered those leads, 25% might be A, 50% B, 25% C. Tier C is who you will never close. The dial number gets hit, but you’ve wasted your time.”

What did Owner do about it? They built what Kyle calls AI PCR - pre-call research - that captures everything between calls automatically. Their outbound BDRs now make 150 to 250 calls a day and stay hyper-prepared throughout. Kyle’s top BDR last month generated $174K in closed-won ARR. Cold. The historical average was $72K per rep per month. They’re now at around $120K per BDR. How? By centralizing AIs.

A lot of teams think the answer is giving every rep their own Claude skill and letting them generate their own lead lists. Kyle says no. One specialist owns the system, builds it to production quality, and deploys one version at scale. Expert vs. non-expert in applied AI isn’t 50% better or 100% better - it’s 20x better. You need the expertise concentrated, not scattered.

Your AEs don’t need to be AI-savvy. This was the most controversial line in his chapter. Kyle wants to encourage AI fluency everywhere, but he’s clear: making your AEs better at AI doesn’t move the needle the way people assume. Fixing the data and the system does.

A lot of that gain didn’t come from better scripts. It came from giving reps only the leads worth calling.

If you’ve been telling yourself that hiring more BDRs is the answer to your pipeline problem, this chapter is going to be uncomfortable reading. In the best way.

What All Three Have in Common

When I finished the Atlas, I noticed something: Elena, Emily, and Kyle are talking about completely different parts of the funnel - capture, qualification, outbound - but they’re saying the same thing.

Stop adding more. Start removing waste.

• Elena: stop minimizing freemium, give away the product so the buyer can actually feel it

• Emily: stop running random acts of marketing, pick fewer bets and make them count

• Kyle: stop cranking dial volume, fix the data so reps only touch leads worth calling

The 2026 GTM AI playbook isn’t about doing more things faster with AI.

It’s about using AI to strip everything that doesn’t compound.

The teams that figure this out first are the ones we’ll be writing about next year.

The full GTM Atlas has six more chapters I haven’t covered here - Nicolas Sharp (Attio’s founder & CEO) on why there is no universal playbook, Roniesha Copeland (Vercel) on building the system before the message, James Pastan (Framer) on activation, Rati Zvirawa (Intercom) on letting AI own (not just assist), Josh Epstein (Coder) on retention as a culture problem, and yes, my chapter on building a GTM brain.

It’s free, non-gated, beautifully designed, and definitely worth your time.

I read it twice. Your turn.

Explore the GTM Atlas 🌍

Talk soon,

Maja

✅ Need ready-to-use GTM assets and AI prompts? Get the 100-Step GTM Checklist with proven website templates, sales decks, landing pages, outbound sequences, LinkedIn post frameworks, email sequences, and 20+ workshops you can immediately run with your team.

📘 New to GTM? Learn fundamentals. Get my best-selling GTM Strategist book that helped 9,500+ companies to go to market with confidence - frameworks and online course included.

📈 My latest course: AI-Powered LinkedIn Growth System teaches the exact system I use to generate 7M+ impressions a year and 70% of my B2B pipeline.

🏅 Are you in charge of GTM and responsible for leading others? Grab the GTM Masterclass (6 hours of training, end-to-end GTM explained on examples, guided workshops) to get your team up and running in no time.

🤝 Want to work together? ⏩ Check out the options and let me know how we can join forces.

This newsletter is sponsored by Miro.

I’ve helped 10,000+ companies build their GTM plan. I know exactly where they get stuck. Every time, it’s the same three things:

• Too many GTM ideas, no way to cut through them

• Everything feels important, so nothing gets prioritized

• Great strategy session, zero follow-through

So I rebuilt my #1 GTM planning workshop inside Miro - with an AI agent “GTM Strategist” Sidekick trained on my book and checklist to power up the process.

I’ve run this workshop more times than I can count. The framework never changed.

What changed is that now you don’t need me in the room.

Take the workshop for a spin

(free access)

Dear GTM Strategist,

In my ideal world, my content team would consist of 8 AI agents that would transform my call recordings into snappy LinkedIn posts and long-form Substack. 🤡

In the real world, I spend 20 hours/week working on content, we have an editorial and design team, and I am still scratching my head about how to speed up our content production with AI without compromising the quality and brand voice.

Claude Code was a big unlock on this journey since I could finally create context once, and then it knew our business, ICP, and branding essentials without dramatic flights in prompts and projects setups that produce various inputs.

Still, I keep hunting for more structured, end-to-end examples of how GTM operators are doing this - and ironically, I found one of the best in the LinkedIn comments.

Benjamin Gibert is the CMO of Base Operations, a B2B SaaS platform selling street-level threat intelligence to corporate security teams and government agencies. It is a series A startup with ~40 people. Ben does not have a dedicated content team. He is building an AI-native GTM strategy using agentic systems, but he can not get away with AI slop in a field that requires high trust.

Yet, his best marketing coworker is Claude Code.

Benjamin has built a content system that takes structured company knowledge, regularly enriches it with real customer conversations, and turns it into SEO articles, LinkedIn posts, ABM personalization, and even product strategy inputs.

Stuff that we like to call content engineering - the post you’re about to read is the most concrete walkthrough of how to actually do it that I’ve come across. Directory structure, agent specs, where the human review gates sit, which integrations matter and which don’t. He’s also open-sourced the whole thing on GitHub, so you can pressure-test it on your own setup.

Everything that we’ll present is done purely with the Claude Code and some power-up tools in the price range of SMEs. In other words, you can do it too!

Benjamin, how can we set up AI content workflows that produce content that is factually true and actually produces pieces that our ICP appreciates, instead of calling them AI slops?

Everyone in GTM is trying to figure out how to build content workflows that drive pipeline and revenue with AI. The problem is that if your AI-generated content is indistinguishable from your competitor’s AI-generated content, it won’t make any difference.

I run marketing for Base Operations, a B2B SaaS platform that provides street-level threat intelligence to corporate security teams. Most of my working hours are inside Claude Code. Over the past 14 months, I’ve learned (mostly through painful trial and error) how to build a content system that takes sales call transcripts and produces SEO/AEO articles, LinkedIn thought leadership posts, ABM personalization inputs, product strategy insights, and ultimately helps refine our targeting and GTM strategy.

Sales calls have always been a goldmine of customer insights. AI gives you a unique opportunity to turn that into the fuel for your GTM engine, but how you set up your system is the difference between sophisticated slop and differentiated content that converts your ideal buyer.

This post walks through how the system works: the knowledge foundation, the intelligence extraction, the pipeline stages, and the tool integrations that close feedback loops. I’ll also share the framework that puts it all in context, and point you to an open-source repo where you can build your own version.

The pipeline is the second thing I built. The first was the knowledge foundation underneath it, which is the context that matters most.

The Sophisticated Slop Problem (And What Solves It)

I’ve seen a lot of B2B teams embrace AI for content and fail. They build elaborate AI workflows that produce output that looks good only on the surface. I’m sure you’ve seen it. It reads well, hits the right keywords, and feels well-researched. And it says absolutely nothing a competitor couldn’t copy-paste, swap in their logo, and publish tomorrow.

I call this “sophisticated slop.” It passes every surface-level quality check, but it ranks nowhere and converts nothing, because search engines and buyers can both tell when content lacks genuine specificity. It erodes trust in humans and it doesn’t get ranked by algorithms, which are quickly adapting to the flood of AI content.

I made this mistake myself early on. My first attempt at a content pipeline had impressive multi-step workflows, carefully written prompt files, and zero strategic or tactical context beyond basic company information and target audience. After over a decade writing B2B content and managing content teams, I was so excited about building the pipeline architecture that I didn’t pause enough to evaluate how differentiated the output actually was.

Then I realized 90% of output quality comes from what you feed the system, not from the sophistication of the agents. This sounds obvious, but it’s easy to miss. Simple agents reading rich, specific context will outperform complex agents reading thin context every time.

A useful gut check: could a competitor copy-paste this output and change the logo? If yes, it’s slop regardless of how impressive the pipeline or how sophisticated the content sounds. That test has everything to do with whether the system contains genuine intelligence about your market, your customers, and your competitive position.

The fix is context engineering: building something substantial before you build the first agent. If you take one thing from this post, it’s that the quality of what you feed the system determines output quality more than any skill architecture. And the act of writing that context (documenting your ICP with real data, articulating positioning based on differentiated value, codifying voice rules with anti-patterns) forces the strategic clarity that makes everything downstream specific.

The Knowledge Foundation

The foundation of the system is a modular context architecture that lives in Github: a structured set of interconnected files that agents consume selectively.

The master knowledgebase is the source of truth. One document containing positioning, ICP definition, persona profiles with buying psychology, voice and style rules, competitive landscape, proof points with metrics, and messaging framework. For Base Operations, this covers everything from our three core points of view to prohibited terminology to how we talk differently to a CSO versus a security analyst.

From the master, I derive scoped module files that agents actually consume:

• icp.md -- firmographics, technographics, tiered customer profiles, buying committee roles, team structure, external signals that indicate a company is in-market

• personas.md -- per-persona buying psychology, pain points in their language, decision criteria, common objections, content preferences, and where they spend time online

• positioning.md -- strategic narrative, points of view and insights that ladder to them, differentiators with proof points, value propositions mapped to specific pains

• voice-guide.md -- tone rules, voice guidance, style and editorial standards, anti-patterns (no “it’s not X, it’s Y”), a list of prohibited terms (no “seamless” or “revolutionary”)

• competitive-landscape.md -- competitor mapping, battlecard data, positioning gaps to exploit

The cascade principle makes this maintainable. When positioning changes, I edit the master knowledgebase, run a sync skill, and every derived module updates. Every agent that reads those modules produces different output on its next run. Maintain in one place, consume in many. Without this, you end up maintaining the same positioning language in 15 different prompt or skill files, and they drift within a week.

Why modular instead of one big file? Each skill specifies exactly which context files it reads. The research agent reads ICP, competitive landscape, and customer intelligence. The writer reads voice guide, positioning, and the approved outline. Tight context scoping produces better output, uses fewer tokens, and reduces the chance of an agent pulling irrelevant information into its reasoning.

The customer intelligence layer is where the real differentiation lives. We’ve extracted hundreds of structured insights from sales call transcripts across six dimensions: jobs to be done, customer problems, workflow reality, buying decisions, competitive intelligence, and product-market fit signals. These are structured JSON with exact customer quotes, speaker context, pain triggers, cost estimates, and urgency signals for every call.

This is the raw material that makes content sound like a practitioner wrote it instead of a marketing team. When a security director at a Fortune 500 retailer says “We have 80,000 employees. Realistically, we have budget to monitor threats for maybe 50 of them. The executives. Everyone else, we just hope nothing happens,” that quote, properly anonymized, does more work than any marketing copy I could write. It names the impossible math that every security leader knows but nobody talks about in vendor conversations.

The intelligence layer also includes a customer language lexicon: the specific terms buyers use versus the terms marketing teams use. In our market, customers say “single pane of glass” while marketers say “end to end platform.” Customers say “I need to know what’s happening on this block” while marketers say “hyperlocal situational awareness.” When your content uses the customer’s language, it feels like a peer wrote it. When it uses marketing language, it feels like a vendor wrote it.

The proof library completes the foundation: case studies with quantifiable metrics (”70% reduction in assessment time,” “5x increase in location coverage”) mapped to specific value propositions. Every claim the system makes in an article is backed by a specific number from a specific customer outcome. Claims without proof are invisible to both search engines and skeptical buyers.

All of these files are in markdown. This is the most effective format for AI agents to search and cross-reference while minimizing token costs. Plain text, version-controlled in git, readable by any agent without special parsing.

Here’s what the directory structure looks like:

A side benefit I didn’t anticipate: once your GTM assets are structured like this in a repo, every file becomes searchable and cross-referenceable by Claude Code. You can ask research questions across your entire customer intelligence base (”Which pain points come up most often in enterprise healthcare calls?” or “What do prospects say about Competitor X’s pricing?”) and get answers grounded in your actual data. The structured repo becomes a research tool in its own right. And you aren’t locked into your call recording vendor’s credit system, pricing, or schema markup.

One Input, Multiple Outputs: The Customer Intelligence Multiplier

This is the strategic idea that changed how I think about content engineering. A single structured extraction from a sales call produces intelligence that feeds every GTM motion.

The call analyzer takes a raw transcript and produces structured JSON. The extraction runs through a Claude Code skill with a detailed system prompt specifying the exact output schema. You drop a raw transcript from any call recording tool (Gong, Fathom, Otter, or even manual notes) and the skill produces the structured JSON. The prompt and example output are both in the repo. The JSON output covers:

• Jobs to be done: what the customer is trying to accomplish and why, with desired outcomes

• Pains with context: specific problems, triggers, costs (time, risk, money), urgency signals

• Workflow reality: current processes, manual steps, spreadsheet hacks, brittle integrations

• Competitor mentions: brands trialed, workarounds used, sentiment toward each

• Customer lexicon: exact phrases customers use versus marketing-speak, including “deviant” terms (the non-obvious ways buyers describe your category)

• Keyword candidates: scored by buyer intent (35%), differentiator fit (20%), SERP weakness (15%), business priority (10%), mini-volume plausibility (10%), and authority fit (10%)

That structured output feeds:

SEO and AEO articles via the content pipeline (the detailed example below). Pain-point keywords, buyer language, and evidence from the call become the raw material for articles that rank because they contain genuine specificity.

LinkedIn thought leadership through a parallel pipeline. The same call produces anonymized customer quotes, tension hooks (”80,000 employees, budget for 50”), and persona-specific framing for founder content.

ABM personalization through persona-specific pain points, industry patterns, and objection mapping that flow into outbound sequences. When an outbound email references the exact workflow problem a prospect’s peer described on a sales call, the response rate is categorically different from a generic insight and ask.

Product and GTM strategy through JTBD pattern analysis, feature request clustering, and competitive intelligence aggregation across calls. When three different security directors independently describe the same gap in their workflow, that’s a signal your product team needs to hear. The structured extraction makes that pattern visible.

There’s also a compounding effect on your strategy itself. Run the call analyzer systematically across all of your sales conversations and patterns emerge that inform your ICP and persona definitions upstream. We discovered that mid-market security teams react to AI in security fundamentally differently from enterprise teams, which led us to split a single persona into two with different content strategies. That insight came from structured analysis of 150+ calls, not from any individual conversation.

A note on cadence: this kind of strategic update to ICP and positioning should happen every 3-6 months based on accumulated evidence, not in reaction to every call. Run the analysis, surface the patterns, then take findings through a structured review with your team. AI surfaces what’s changing; humans decide what to integrate.

One conversation with multiple structured outputs. Each feeding a different system that consumes only the dimensions it needs. The call analyzer produces comprehensive structured intelligence, and each consuming system takes the slices it needs. Before building this, I was treating each GTM motion as a separate content creation problem. Reframing it as an intelligence distribution problem improved the compounding impact of the whole system.

The Content Pipeline: Stage by Stage

Here’s the SEO/AEO pipeline as one detailed example. This is the system that produces differentiated articles that rank, grounded in sales call intelligence no other company has.

The system can produce articles as fast as you can review them. The bottleneck is the human in the loop at the outline stage and final review, which can take at least 15-20 minutes per article. The results have been meaningful: organic traffic doubled over 3-4 months, AEO citations increased 500-600%, and we’ve ranked #1 on Google for core terms within weeks of publishing, resulting in inbound increasingly driving more demo calls and pipeline.

Research and Intelligence Layer

The pipeline starts before any writing happens, with two agents doing the strategic work.

The Call Analyzer (built on Grow & Convert’s Pain-Point SEO methodology) processes a transcript and produces the structured JSON I described above. It prioritizes buyer intent over search volume, captures exact customer phrasing rather than generic summaries, and maps every finding to specific keyword types: category searches, comparison queries, JTBD queries, and deviant terms.

The Research Agent takes the call analyzer output across hundreds of calls and produces two things: a prioritized topic backlog scored by pain-point SEO criteria, and detailed content briefs for selected topics.

The brief is the part that tends to surprise people. It’s a structured JSON document specifying everything a downstream agent needs:

{

  “meta”: {

    “topic_id”: “GC-202501-security-platform-guide”,

    “priority_score”: 4.3,

    “stage”: “BOFU”,

    “format”: “guide”

  },

  “keyword”: {

    “primary”: “security intelligence platform”,

    “variants”: [”threat intelligence software”, “risk assessment platform”],

    “searcher_intent”: “Category”

  },

  “audience”: {

    “persona”: “Global Security Director”,

    “pain_summary”: “Spending 20+ hours manually gathering threat data”

  },

  “value_props_to_weave”: [

    {

      “pain”: “Manual research bottleneck”,

      “prop”: “5x faster threat assessments”,

      “proof”: “Case Study: Fortune 500 Security Team”

    }

  ],

  “originality_nuggets”: [

    {

      “type”: “quote”,

      “source”: “Sales transcript - Fortune 500 retail”,

      “owner”: “Content team”

    }

  ]

}

The brief specifies the keyword, the searcher’s intent, the target persona, the specific angle, a section-by-section outline, which value propositions to weave in (mapped as pain to proposition to proof), originality requirements, internal link targets, and SEO metadata.

The research agent consumes icp.md, personas.md, competitive-landscape.md, and the customer intelligence directory. It never sees the voice guide or the full knowledgebase. It gets exactly the context it needs for strategic research decisions and nothing else.

Enrichment and Outline

Stage 1 (Enrichment): The brief gets enriched by mining five tiers of context sources in priority order:

1. Structured JSON insights (the fast path: already tagged by persona and use case)

2. Case studies (metrics, customer outcomes, before/after comparisons)

3. Use cases (workflow demonstrations, product capabilities)

4. CSV insights database (pattern-level evidence across multiple calls)

5. Full transcripts (deep path, last resort for specific unfilled gaps)

Evidence gets mapped to each section of the brief with specific quotes, metrics, and source attribution. The actual quote, the actual number, the actual speaker role and industry context.

Stage 2 (Outline): The enrichment output merges with the brief structure into a writer-ready outline. Each section includes word count targets, specific proof points assigned with full text, AEO formatting directives (definition boxes, comparison tables, numbered processes), and entities to mention.

Then the pipeline stops.

This is deliberate. The human review gate sits between research and execution. You review the outline, adjust the emphasis, approve or reject evidence choices, and rename the file to _outline-approved.md before the pipeline continues. Human judgment at the strategic inflection point, not at the copy-editing stage where it’s too late to change direction.

Write, Edit, Link, Publish

Stage 3 (Writer): Writes from the approved outline, consuming only voice-guide.md, positioning.md, and the outline itself. The outline carries all the evidence forward, so the writer agent never needs to search the knowledgebase. The outline is the information boundary. Everything the writer needs is already in it. No searching, no improvising, no pulling in context that wasn’t approved.

The writer follows specific patterns. Feature to Benefit to Proof: mention a capability, explain why it matters to the reader, prove it with a specific metric or customer outcome. Value propositions as undercurrent: threaded throughout the piece at natural moments, never isolated into a “Why Our Product” section. Customer anonymization applies during writing: “Bob Rendle, Director of Crisis Management at Compliunce” becomes “the Director of Crisis Management at a Fortune 500 insurance and risk management company.” Keep the title, the industry, the company scale. Remove the name and the brand (unless you can get approval).

Stage 4 (Editor): Style enforcement, terminology standards, prohibited language removal. Reads only voice-guide.md. Catches any remaining drift from brand voice, eliminates weak qualifiers, scrubs obvious AI-writing tells, and ensures consistent terminology. If Stage 3 did its job, Stage 4 makes minor adjustments.

Here’s where another compounding loop lives. After every human editorial pass, everything is tracked in git. A Claude skill tracks the diff between the AI draft and my edited version, and the system analyzes what I changed: which phrasings I rewrote, which claims I softened, which sections I restructured. Those patterns get fed back into the agent’s prompt and workflow rules. The editor that produces draft 15 has internalized lessons from my edits on drafts 1 through 14. This is slow, unglamorous improvement, and it’s the kind of thing that adds up over months.

Stage 5 (Internal Linker): Calculates a link budget based on word count (an 800-1,500 word article gets 5-8 links; a 2,500+ word piece gets 10-15), then distributes priority URLs across the article. Core product pages get priority, then case studies, then other articles. Reads only the sitemap. Hard rules: max one link per paragraph, no duplicate destinations, anchor text must already exist naturally in the prose.

Stage 6 (Publisher): Markdown to HTML conversion, FAQ JSON-LD schema generation for rich snippets, category detection based on content analysis, related article suggestions via keyword matching against the sitemap, and Webflow CMS draft creation. Reads only the SEO assets YAML file generated back in Stage 2.

Each stage loads the minimum context it needs. The pipeline architecture enforces this: no stage has access to more than it requires. This prevents context pollution, where an agent with too much information starts making connections the outline didn’t authorize. I learned this the hard way when an early version of the writer agent had access to the full knowledgebase and kept inserting product mentions that the outline hadn’t approved that didn’t make sense. Everyone who has built with AI knows this pain. Constraining context solved it overnight.

Where Tool Integrations Close the Loop

The pipeline works with Claude Code and local files. No external integrations are required. But the system gets meaningfully more powerful when you wire in data sources through MCP (Model Context Protocol) or CLI (Command Line Interface).

Firecrawl gives the research agent automated SERP analysis. Instead of guessing what’s ranking for a target keyword, it can crawl the top results, identify content gaps, and build briefs informed by what’s actually out there. The enrichment stage can pull competitor content analysis into the evidence map.

HubSpot connects deal data to content prioritization. Which pain points correlate with deals that closed won? Which personas engage with existing content? This turns topic selection from educated guessing into one informed by the fastest closing deals in our CRM.

Clay enriches companies mentioned in transcripts to surface industry patterns across the customer base. When you see that the same workflow problem appears across retail, financial services, and healthcare, you know it’s worth a dedicated article rather than a throwaway mention.

Lemlist (or any outbound sequencer with an API) connects the intelligence layer directly to outbound execution. The same persona pain points, customer language, and objection mapping that feed your content pipeline can also draft multi-channel outbound sequences. Combine structured call insights with an outbound email skill that knows your voice rules and you can go from “interesting pattern in our sales calls” to “personalized 3-step sequence loaded in the sequencer” in a single session. The intelligence does double duty.

Webflow is already in the pipeline. Stage 6 creates a CMS draft directly, with FAQ schema, category detection, and related article suggestions. No copy-paste or manual formatting. The human in the loop is for editorial standards, not copy-pasting markup into your CMS.

Analytics feedback is the loop that makes the system compound. Which articles drive pipeline? Which briefs produce content that ranks? Which topics drive demo requests versus bounces? That performance data flows back into the research knowledgebase, updating priority scores and informing future topic selection. The research agent building brief number 120 has access to performance data from briefs 1 through 119. It knows which angles worked, which personas engaged, which content formats converted. This is where the system starts to learn, and it’s the difference between a content production line and content infrastructure.

The core pipeline works today with Claude Code and local files. Every MCP integration is an optional enhancement. You don’t need all of them to start. You add them as your system matures and the feedback loops become worth closing.

The Bigger Picture: From One Motion to a Compounding System

The content pipeline is one motion in a larger system. I want to zoom out briefly because the architecture pattern transfers to other GTM motions.

I’ve come to think about AI-native GTM as four layers:

Layer 1: Human Strategy and Creative Direction. The modular context files I described. Positioning, ICP, personas, voice guide, competitive landscape. Written by humans, consumed by AI. No agent replaces the judgment that goes into sharp positioning.

Layer 2: AI as Thought Partner. Using AI to pressure-test strategy, explore alternative angles, deepen thinking, test against persona councils. Augmenting human reasoning, not replacing it.

Layer 3: AI-Powered Systems. The content pipeline lives here. So does signal-based outbound, demand gen campaigns, customer intelligence extraction, competitive monitoring, and lifecycle campaigns. Each motion follows the same pattern: signals trigger workflows, workflows produce outputs, quality gates enforce standards.

Layer 4: The Practical Architecture. Claude Code, MCP integrations, skill files, the directory structure that makes it all work. The architecture compounds: every reference file makes every skill more effective. The marginal cost of the 20th skill is far lower than the first.

The learning loop connects execution back to strategy. Content performance data updates the research knowledge base. Outbound response rates refine persona models. Customer intelligence from new calls enriches the foundation. Feedback flows upstream, and the system gets better with each cycle.

The critical insight at every transition between layers: the jump is never about discovering a better tool. It’s about restructuring how work is represented. From conversations to files. From files to skills. From skills to pipelines with learning loops.

One thing that makes the system genuinely extensible is that clear structure in the repo means adding new inputs is straightforward. When I added dedicated competitive intelligence pages for each of our main competitors, every skill that consumed competitive-landscape.md immediately had richer context. That opened the door to competitor-specific content pipelines (comparison articles, alternative pages) that would have required a separate research effort before. The structured foundation means new capabilities often emerge from adding a new file to the right directory.

A helpful way to think about where you are today:

• Level 0 (Ad-hoc): Copy-paste into AI chat. Re-explain your company every session.

• Level 1 (Documented): ICP, positioning, voice rules exist as files AI can read in Claude projects or Custom GPTs.

• Level 2 (Assisted): AI drafts following your strategy. Skills extract insights, write articles.

• Level 3 (Systematic): Signal-to-workflow-to-output patterns for each GTM motion.

• Level 4 (Compounding): Performance data updates strategy. The system learns from every execution cycle.

The content pipeline I walked through operates at Level 3, moving toward Level 4 as the analytics feedback loop matures. The jump from Level 0 to Level 1 is the hardest, and also the most valuable even if you stop there. Writing a detailed ICP document for an AI to consume forces you to actually have a detailed ICP. Articulating positioning with proof points for a skill file to reference forces you to actually have sharp positioning. The foundation-building work makes your GTM better whether or not you ever build the pipeline on top of it.

Try It Yourself

I’ve open-sourced a generalized version of this system as a GitHub repository. It includes:

• Example files from a fictional B2B SaaS company showing what good looks like at each stage (fully written examples you can study, not placeholders)

• Strategy templates that Claude Code walks you through interactively to build your knowledge foundation

• Execution skills: call analyzer, research agent, the full 6-stage SEO pipeline, LinkedIn insight extraction

• A sync skill that cascades master knowledge base edits to all derived module files

• An onboarding agent (via CLAUDE.md) that detects your setup state and guides you through building your own version

Clone it. Open Claude Code. The system walks you through creating your knowledge foundation first, then running the pipeline.

Get the repo by Subscribing to Dirty Hands.

The repo is designed to grow. The content pipeline is the first fully built motion I’m offering for free.

I get asked for advice on AI all the time, so I’m launching the Dirty Hands GTM newsletter to share what I’m building for GTM (and selfishly to force myself to clarify my thinking on this). It’s one of the wildest and most exciting times to be in B2B marketing if you love getting your hands dirty and building. It’s a cliché because it’s true.

Every week, I’ll break down one system like this in Dirty Hands, a newsletter for GTM operators who build with AI. Real systems with catalogued results and failures.

Future issues of the newsletter will add new motions to the repo: signal-based outbound, content repurposing, customer intelligence extraction, competitive monitoring. Each new motion adds skills, so you have a reason to keep coming back as the system expands.

If this walkthrough was useful, the newsletter goes deeper.

Subscribe to Dirty Hands on Substack

✅ Need ready-to-use GTM assets and AI prompts? Get the 100-Step GTM Checklist with proven website templates, sales decks, landing pages, outbound sequences, LinkedIn post frameworks, email sequences, and 20+ workshops you can immediately run with your team.

📘 New to GTM? Learn fundamentals. Get my best-selling GTM Strategist book that helped 9,500+ companies to go to market with confidence - frameworks and online course included.

📈 My latest course: AI-Powered LinkedIn Growth System teaches the exact system I use to generate 7M+ impressions a year and 70% of my B2B pipeline.

🏅 Are you in charge of GTM and responsible for leading others? Grab the GTM Masterclass (6 hours of training, end-to-end GTM explained on examples, guided workshops) to get your team up and running in no time.

🤝 Want to work together? ⏩ Check out the options and let me know how we can join forces.

This newsletter is kindly supported by Attio - the AI-native CRM built for modern GTM teams. If you’ve ever battled a rigid legacy CRM, Attio fixes that with clean data, real-time visibility, and workflows that actually speed you up.

Today, 5,000+ fast-moving companies rely on Attio’s flexible data model, modern design, and builder-friendly setup to shape their CRM around how they work - not the other way around. It’s built for founders, revenue ops, and GTM teams who need a system that adapts as fast as we do.

So if 2026 is the year you finally break up with your legacy CRM or start fresh on a new project, this is the perfect moment to take Attio for a spin.

Try Attio for free

Dear GTM Strategist,

Claude mania has taken the GTM AI space by storm in Q1 2026. People who have never written a line of code are now spending their nights and weekends staring at a console, shipping GTM systems that used to require a full engineering sprint.

And Anthropic, the company behind Claude, just crossed $30B in ARR - in record time:

But is the hype real? Are GTM operators actually getting value from all their cool builds, or are we entering another “crypto” era - lots of talk, energy, questionable ROI?

Time to find out.

Kyle Poyar (Growth Unhinged) and I surveyed 200 GTM AI operators and curated their best insights into the 2026 Claude for GTM Pulse Report, focusing on how they use Claude Cowork and Code, the value teams get from these tools, and their limitations.

In this article, I am presenting my 5 key takeaways from a 37-Page report:

1. Claude Cowork is not (just) Claude Code for Dummies. Claude Code and Cowork split GTM operators almost 50/50, but they are not complete substitutes.

2. 80% use Claude for productivity. Content creation (29%) and GTM engines (23%) deliver the most impact at the moment, but it feels like we are just scratching the surface of what is possible.

3. GTM engineers are wiring Claude Code into stacks and systems, not using it solo but in combination with other tools. Most common use cases so far: 79% are building SDR research tools and 71% adapting messaging dynamically by segment.

4. 92% reported that Claude Code and Cowork help them “save time”, but more interestingly, 67% have reported building workflows that were previously impossible. But just 27% have replaced a GTM tool so far - the SaaS apocalypse crowd might be getting ahead of the data.

5. But nothing is perfect - even Claudes. Anthropic is a pretty savvy monetizer - even Max plan ($100-$200/month) users hit credit walls 2-3x a day. Memory loss, integration gaps, and a lack of team skill-sharing round out the list of frustrations. We’ve also gathered some intel on how best-in-class users bypass these limitations.

A massive thank you to every community member who participated in the survey 🙏 - this article would not have been possible without you. You are the best!

Want more insights?

Here you go.

Insight 1: Is Cowork Just Claude Code for Dummies?

When Anthropic introduced Cowork in January 2026, my first reaction was: “OK, this is Claude Code for people who don’t want to open the terminal.” I was somehow wrong. We went ahead and bypassed Cowork to go all in on Claude Code, but after talking to colleagues who built legitimate, serious use cases in Cowork, I revised my opinion.

For example, one of the well-known Substackers in my circles let go of three VAs because Cowork could just absorb their workload for content research and repurposing.

Cowork is not just a lite version of Claude Code - it is a different tool optimized for a different kind of work, and it feels like they are just getting started, so we should not dismiss them as some “inferior version” of Code.

Our data backs this up: just three months after launch, Cowork (32%) has already surpassed Claude Code (31%) as the primary Claude product among GTM operators we surveyed.

Claude Chat sits at 30%, and many of us use all tools at the same time, which tells you something important: there is no “one to rule them all” - they do different things in different time and cost differently - here is a quick breakdown on when to use each.

A reality check: this survey sample skews toward advanced users, so the share of Claude Chat users in the broader population is much higher. Don’t let the numbers intimidate you - use them as inspiration instead.

The core insight from the data: 93% of Claude Code users rely on detailed company context1 as a core foundation, vs. 72% of Cowork users. That gap tells you who is building a compounding system and who is running sophisticated tasks.

For advanced users: The other thing where Claude Code outshines its siblings is interconnectivity with other tools for complex use cases and GTM systems. If we are somehow limited with native apps and MCPs within chat and Cowork, CLIs and API integrations fuel powerful, customizable Claude Code builds at optimized costs. Bring agents into the mix and you can unlock the kind of agentic AI systems Canva, Descript, and Linear have built2.

Insight 2: Most Teams Are Still in Productivity Mode - but #1 Use Case is Content

The most widely adopted use case is productivity at 80% - analyzing data, preparing for meetings, and creating documents. Content creation comes in second at 69%, followed by product marketing at 64%.

That is the adoption story. But here is the more interesting cut: which use case do people say has the biggest impact?

Content creation jumps to #1 impact at 29%, while productivity drops to 15% despite leading on adoption. GTM engines punch way above their adoption weight, cited as the #1 impact use case by 23% of respondents despite only 54% adoption.

Long story short: people use Claude most for productivity, but they get the most value from content and GTM engines. Specifically, Claude Cowork shines for content creation, while for Claude Code users, GTM engines are the primary use case.

Within content, the top workflows are:

1. Creating social media posts (72% of content users)

2. Building websites, landing pages, and micro-sites (66%)

3. Creating structured content pipelines (52%)

4. Auditing SEO performance (38%)

5. Analyzing traffic and content performance (34%)

The thing that separates content in Claude Code/Cowork from content in Claude Chat is context engineering. With Code or Cowork, you can encode your brand voice, your ICP, your top-performing posts, your sponsor briefs - and Claude pulls exactly what is relevant for the task at hand. The more you feed the system, the better the output gets.

But remember - with AI content creation, it is often “garbage in, garbage exponentially out”. Strategic narratives and unique points of view are more important than ever:

Without something original to say, the best architecture in the world just produces polished noise. The system amplifies your point of view. It does not create one for you. - Amos Bar-Joseph, CEO at Swan AI

To help make LinkedIn a better place for the intelligent carbon life still being there, I am sharing a core idea of how we structured the content system in our Claude Code. It is already producing amazing LinkedIn pieces (still human edited!) - design briefs are OK but need final touch and AEO/SEO. Well, we still have room for improvement there - so if you have some advanced builds, send them my way, and I’ll consider sharing them in our future content.

Now back to survey results:

Insight 3: GTM Engine Builders Are Playing a Completely Different Game

The 54% of operators using Claude for GTM engines are not just doing things faster. They are building workflows that did not exist before. Here is what the top GTM engine use cases look like:

What makes this category different is the combination of tools. GTM engine builders are not using Claude in isolation. They are wiring it into a stack. The tools that Code and Cowork users love the most are ChatGPT (#1, mentioned by 26% of respondents), Clay (#2 at 21%), HubSpot (#3 at 19%), LinkedIn (#4), and Apollo (#5).

The top GTM AI tool stack in 2026 is Claude + CRM + orchestration. These tools are not competing. They are combining.

Claude handles the intelligence layer. Clay handles data enrichment and workflow automation. It connects the CRM layer. LinkedIn and Apollo feed the signals. The builders who get this are not asking “which AI tool should I use?” They are asking “how do I architect a system where Claude sits at the center, reads my context, connects to my stack, and runs while I sleep?”

Insight 4: Show me the money - what value are users getting from Claude

Every reasonable human being at this point feels like - OK, sounds like an interesting opportunity, but it does take a lot of work and changes. Is it worth the trouble or are GTM nerds just raving to each other to have something new to say? In other words - is the hype real?

Respondents are practically unanimous (92%) that Claude Code and Cowork have saved them time. But what really sparked my attention is that 67% said that it enabled them to do something previously impossible. Enabling previously impossible workflows is a different category of value - that is where compounding GTM intelligence lives.

What’s really interesting is whether Claude replaces existing tools. 55% of respondents said yes, but the breakdown was more intriguing.

Lower qualified work and non-core contractors have taken the larger hit so far 😢 - remember the Substack guy who could fire 3 VAs (yeah, that is the spiel ATM)? While having a good content person is invaluable, it is mission-critical to learn AI for the lower-qualified tasks. I’ve been saying that to every young person out there: if you cannot compete on quality yet, compete on speed and productivity - make AI your bestie.

On the tool replacement side, only 27% have replaced an existing GTM tool so far, 30% say they will soon, and 43% say no. Most at risk are contractors, SEO agencies, ad agencies, spreadsheets, manual outbound work, ChatGPT, Perplexity, analytics providers, and third-party automation tools.

Some are loudly predicting a SaaS apocalypse driven by vibe coding. Considering previous findings, this is not exactly true. I personally believe in combining the tools with strong ecosystems into new stacks that empower GTM systems.

Insight 5: The Frustrations Are Real, Documented, and Partially Fixable

This section exists because Emily Kramer (MKT1 Newsletter) suggested we ask about frustrations when she reviewed our survey. Best suggestion we got. The responses were juicy.

The reported limitations of Claude, ranked by number of mentions:

1. Credits, cost, and usage limits - running out of credits, throttling, usage limits even on the Max plan, unpredictable usage on small tasks.

2. Memory limitations - losing context between sessions, no persistent long-term memory, manual workarounds required.

3. Context window and token limits - running out mid-task, large operations failing, duplication across sessions.

4. Integration and data gaps - weak Google and Salesforce integrations, limited connectors, restricted access to company data sources.

5. Reliability and consistency - hallucinations, shallow outputs, inconsistent results across sessions.

6. Team collaboration - skills are not easily shareable, updating a skill does not cascade to teammates, poor team workflow support.

“I’m on a Max plan and I still run into model usage restrictions 2-3x per day.”

“If a coworker comes up with a skill that’s useful, we’ll have to build the skills separately rather than sharing.”

Some practical “hacks” that help:

• Open new chats regularly instead of continuing long threads - reduces context bloat

• Ask Claude to plan before executing - use plan mode before burning tokens

• Move from MCP to CLI or direct API calls for reliability - ScaleKit benchmark data shows CLI hits 100% reliability vs around 72% for MCP, at a fraction of the token cost

• Mix Sonnet model into your workflows instead of Opus for simpler tasks - it is about 20% of the Opus cost

• Build an external memory system - a CLAUDE.md or history.md file Claude updates proactively

• Test on a small sample before running full campaigns - do not burn credits on untested logic

• To improve output quality, ask Claude to critique itself before you accept the output

“The biggest trap with Claude (or any LLM) is that it’s a yes-man by default. You bring it a plan and it’ll tell you it’s brilliant, maybe suggest a few tweaks around the edges. That’s not useful. I started building a habit: create the plan in Claude, then explicitly ask it to red team the thing.” - Andy Nester, CMO at PERSUIT

In the Future, Everybody Will Be an Engineer

In 1968, Andy Warhol famously said, “In the future, everyone will be world-famous for 15 minutes.”

In GTM AI in 2026, the equivalent might be: in Q4 2025, everybody became a ______ engineer (place your buzzword of choice to fill in the blank).

The trend started with the GTM engineer role, then the term content engineering started trending; oh, and have you ever heard of vibe coding? And just last week I heard about a marketing engineer for the very first time. It’s getting really hard to catch up with all the new engineers - it used to be easy (front-end/back-end). 🙈

It’s an incredibly exciting time to be building in this space and if these strategic narratives usually get great buzz on social media - so by all means, do it - but is all these engineering making non-technical people anxious, excluding women (lower representation in the engineering skill pool) and turning off competent operators who actually know their s*** and favoring tool jockeys instead?

Maybe …

But there are a lot of great things going on in the space, too. There are great companies, teams, solo contributors and content creators pushing the space forward. And sometimes that means having a critically acclaimed discussion or open-sourcing something outside the comfort zone.

I feel we are just getting started here and that Claude’s domination will continue throughout 2026. 🤞 So please keep track of what is new in the space and share your best practices.

The gap between AI power users and those who use GPT chat to prompt outbound emails and blog posts is getting wider and wider, and while you are still early in the space, the AI-powered teams are a force to be reckoned with, not something to smirk at.

Kyle and I hope this report is a high-quality contribution to what is actually happening with Claude in GTM - not just another hype piece. If it was useful, hit that heart on Substack - it helps me convince Kyle to do more of these collaborations 😉.

The full 37-page “Claude for GTM Pulse Report” is available to premium subscribers of Kyle Poyar’s newsletter. It includes more survey data and insights on Claude Code and Cowork adoption, how top GTM teams use Claude, limitations to be aware of, and advice to get the most out of Claude.

Get it here

And if you like what I do and want to support my free content stream, you can do that by buying my digital solutions 👇

✅ Need ready-to-use GTM assets and AI prompts? Get the 100-Step GTM Checklist with proven website templates, sales decks, landing pages, outbound sequences, LinkedIn post frameworks, email sequences, and 20+ workshops you can immediately run with your team.

📘 New to GTM? Learn fundamentals. Get my best-selling GTM Strategist book that helped 9,500+ companies to go to market with confidence - frameworks and online course included.

📈 My latest course: AI-Powered LinkedIn Growth System teaches the exact system I use to generate 7M+ impressions a year and 70% of my B2B pipeline.

🏅 Are you in charge of GTM and responsible for leading others? Grab the GTM Masterclass (6 hours of training, end-to-end GTM explained on examples, guided workshops) to get your team up and running in no time.

🤝 Want to work together? ⏩ Check out the options and let me know how we can join forces.

Most B2B companies spend months optimizing SEO/AEO for one language while leaving 95% of the world’s search traffic on the table. And now, with AI search pulling from indexed multilingual content, the gap is only getting wider. Weglot translates your site in minutes, generates proper language-specific URLs, and handles technical SEO automatically - so you show up in both traditional search and AI-powered answer engines across every market.

Try Weglot free

Use the code MAJAXWEGLOT3 for 3 months of free access on any plan.

Dear GTM Strategist,

Here’s a question that should keep every B2B founder up at night: What does AI say when someone asks for the best tool in your category?

The new “homepage” for your buyers is no longer your website. It’s a ChatGPT prompt. And if you’re not in the answer, you don’t exist.

ChatGPT, Perplexity, and Google AI Overviews are the new front doors for buyer research. According to Forrester, 94% of B2B buyers now use AI during their purchasing process — and most companies have no idea what’s being said about them (or worse: nothing is being said at all).

Prospects visit your website later in the buying process, but are better informed.

So it’s no surprise that 51% of GTM practitioners said they plan to increase their investment in AI discovery and Answer Engine Optimization (AEO) in 2026 (source) — the top channel bet by a wide margin.

But the playbooks are still being written in real time. It’s part gold rush, part black box, and part “nobody actually knows what they’re doing yet.”

So I rolled up my sleeves and did the research on how to do AEO in 2026 - for B2B products. This article gives you:

1. Why AEO is the most urgent GTM priority right now — the data that makes this undeniable

2. What AEO actually is (and what it’s not) — cut through the naming chaos and snake oil

3. A 6-step playbook to audit and improve your AI visibility starting this week

4. Real case studies from companies getting 10-25% of signups from AI discovery

5. The tools and stack to monitor your “Share of Model” and track AI-referred pipeline

Let’s get into it.

The shift is no longer theoretical. It is measurable, accelerating, and directly impacting pipeline.

AI search is massive — and growing fast. ChatGPT now has 800 million weekly active users processing 2.5 billion prompts per day. Google AI Overviews appear in 13% of all searches — double since January. Gartner predicts that by end of the year, 25% of organic search traffic will shift to AI chatbots and virtual assistants.

The conversion quality is insane. This is what really caught my attention. Webflow reported that their ChatGPT traffic converts at 24% — 6x higher than Google. Ahrefs found that AI search accounts for just 0.5% of their visitors but 12.1% of signups — a 23x higher conversion rate. Surfer SEO sees roughly 25% of new customers coming from AI assistants.

AI-referred traffic is a fraction of the volume but disproportionately high-intent. These buyers show up pre-sold because the AI already told them you’re the best option.

And yet almost everyone still feels in the dark when it comes to optimizing for AI search. A 10Fold study of 400 senior marketing execs found that only 11% of B2B firms have the majority of their content ready for AI discovery. There’s a massive gap between awareness and execution — and that gap is your window.

What AEO Actually Is

AEO (Answer Engine Optimization) is about making your product the answer when buyers ask AI for recommendations. Not ranking on a results page. Not getting clicks. It’s about being cited, summarized, and recommended inside a conversation.

The naming situation is a mess - I’m sticking with AEO, but it’s also called GEO (really? 🌍), LLMO, AIO (all in one?), SXO, AI SEO… But the name matters less than the concept.

Here’s the mental model shift that matters:

In SEO, you compete for rankings. In AEO, you compete for mentions. The new metric isn’t “Position 1” — it’s “Share of Model,” which measures how often your brand is the recommended answer across AI platforms. Webflow tracks this obsessively and has grown their AI category share from 1.2% to roughly 60%.

What AEO is not:

• It’s not “new SEO.” AEO extends SEO — it doesn’t replace it. Late-funnel SEO is the foundation. If you don’t rank for “[your category] for [use case],” you probably won’t show up in AI answers either. AI models pull from the same sources that rank in traditional search.

• It’s not gaming LLMs. There’s a lot of snake oil out there — prompt injection tricks, hidden text, keyword stuffing for AI. This is short-term thinking at best and will get you flagged at worst.

• It’s not just a content project. This is what Kyle Poyar warns about most: “AEO isn’t a content thing or an SEO experiment. It’s a GTM capability.” It touches product marketing (how your product is described), content (what you publish), and brand (what third parties say about you).

• It’s not about publishing more. Volume without structure and authority doesn’t improve AI visibility.

The AEO Playbook: How to Get AI to Recommend You

Here’s what I’d recommend based on what the fastest-moving companies are actually doing — not theory, but deployed and producing results.

Step 1: Audit Your AI Visibility

Before you optimize anything, you need to know where you stand. This takes 30 minutes and will probably surprise you.

Run the prompts your buyers would run. Open ChatGPT, Perplexity, Gemini, and Google AI Mode. Type the questions your ICP actually asks: “What’s the best [your category] tool?” “Compare [you] vs [competitor].” “Best [your category] for [specific use case].” Do this for 15-25 variations across the buying funnel.

Track three things for each prompt:

• Are you mentioned at all?

• Are you recommended or just listed?

• How are you described? Is it accurate? Is it how you’d want to be positioned?

You might be the #1 result on Google and completely invisible in ChatGPT. These are different systems with different signals. One company can dominate in ChatGPT and be absent from Perplexity — each platform has its own retrieval logic and citation patterns.

Tools that can help: HubSpot’s Search Grader (free baseline), Profound, Xfunnel, Ahrefs Brand Radar. But for the start, manual testing will do. There’s no substitute for seeing exactly what AI says about you right now.

Step 2: Build Your Prompt Library

Once you’ve done the initial audit, systematize it. Identify the 15-25 core prompts that define your category across the buying funnel:

Top of funnel: “What is [category]?” “How does [category] work?” “Do I need a [category] tool?”

Mid-funnel: “Compare [you] vs [competitor].” “Best [category] for [industry/use case].” “What are the pros and cons of [you]?”

Bottom of funnel: “[You] pricing.” “[You] vs [competitor] for enterprise.” “Is [you] worth it?”

Map yourself and each competitor against every prompt. This becomes your AEO scoreboard — your “Share of Model” tracker. Update it monthly because AI outputs shift with model updates and new content entering the training data.

This is not unlike building an ICP or a messaging framework — you’re defining the battleground. Except instead of a sales pitch, you’re defining the conversations AI is having about your market without you in the room.

Choosing the right prompts: a view from the practice

Not all AI searches are equal, says José Velez, co-founder and CEO of Reach, an AI search research lab running R&D on how AI search engines actually decide what to recommend. “The ones that matter are the high-intent queries across your buyer journey: problems your ICP is searching for solutions to, the specific solutions and categories they're evaluating, and the comparisons they're running between specific vendors. Volume metrics from traditional SEO don't apply here; what matters is intent and proximity to a buying decision.”

So after preparing your list of core prompts, you should reverse-engineer what’s actually in the answers. “For each target prompt, study what the models are citing in that case. Is it a listicle? Owned content from vendors? Review sites? What categories of companies are included? What comparison factors matter? This tells you exactly what your content must cover, what format it should take, and whether you can realistically influence the answer.”

Step 3: Win at Late-Funnel SEO First

Here’s the thing that should relieve a lot of anxiety: you don’t need to start from scratch. Late-funnel SEO is the foundation for great AEO.

AI models rely on live web search (through RAG — retrieval-augmented generation) to construct their answers. They pull from the same content that ranks in traditional search. If your best comparison pages, use case pages, and product pages rank well for high-intent keywords, you’re already ahead of most competitors for AEO too.

What to prioritize:

• Comparison pages: “[You] vs [Competitor]” — these are gold for both SEO and AEO

• Use case pages: “Best [category] for [specific industry/team/problem]” with specific, citable details

• Product pages with clear definitions: Open with what you are, not marketing fluff. AI pulls the first clear definition it finds

• ROI/pricing pages: Detailed, transparent, and structured — these directly feed AI responses about your product

• Technical docs and integration guides. This one is wildly underrated.

Talal Syed leads AEO and SEO strategy at GrowthX. They ran a study across 1 million AI responses and found something that should change how teams prioritize content:

“Docs pages - like technical docs and integration guides - were 8.6x more likely to get cited than other pages on a site when buyers were actively evaluating products. The long tail has exploded with AI search, and these engines are hungry for high-quality content that goes deep, is well organized, and answers the specific questions people have about products. The brands filling that gap on their own sites are the ones showing up in AI.”

If you’re a B2B SaaS company and your docs are an afterthought, you’re leaving citations (and pipeline) on the table.

For product companies, José Velez consistently sees the same category of underproduced, high-intent content driving outsized results: competitor pricing guides, competitor reviews tied to specific use cases, jobs-to-be-done alternatives (”How to embed tables in Notion” - with your product positioned as the faster option), and “How to cancel [Competitor]” articles that intercept buyers already looking to switch.

“These capture buyers at the exact moment they’re evaluating alternatives, which is exactly what AI models surface as answers, but most teams don’t produce them because they’re not intuitive. Disclaimer: don’t apply this blindly without first validating there’s real search demand in your space,” says José.

Step 4: Structure Content for AI Interpretability

AI systems don’t read content the way humans do. They extract, chunk, and reassemble. Your content needs to be built for this.

Freshness is critical. Research from AirOps found that 83% of AI citations come from pages updated within the past 12 months, with more than 60% refreshed within six months. If your best content is a year old and untouched, it’s losing AI visibility every quarter.

Practical formatting changes that matter:

• Lead with direct, definitional answers. Don’t warm up for three paragraphs before defining what your product does. AI pulls the first clean answer it finds. Start every key page with a single, citable sentence: “[Product] is a [category] that does [specific value].”

• Use clear, semantic headings that mirror the questions buyers ask. Not “Our Approach” but “How [Product] Handles [Specific Problem].”

• Modular content structure. Each section should stand alone when extracted from the page. AI systems break content into passages, not pages.

• Add FAQ sections with schema markup. These are directly extractable by AI. Use real customer questions, not marketing-speak.

• Implement structured data (Schema.org) — Product, FAQ, HowTo, Organization schemas make your content machine-readable. This isn’t optional anymore.

Case Study: Webflow’s FAQ + Schema Markup Automation

Josh Grant, previously VP of Growth at Webflow, shared how a focused experiment on just six core feature pages (Design, CMS, SEO, Shared Libraries, Interactions, Hosting) drove outsized AEO results - with no content overhaul and no agency.

His team built an AI workflow in AirOps that used Perplexity to deep-research Google’s “People Also Ask” results, Reddit threads, and niche forums to surface the exact questions buyers were actually asking. From there, the workflow identified gaps in existing FAQ content, generated high-intent Q&A pairs, wrote on-brand answers, and auto-structured everything into clean schema markup.

The results: +331 new AI citations (57% of all new citations across Webflow.com that period) and +149K SEO impressions, a 24% lift. As Josh put it: “AI didn’t change the game here. It just raised the bar for how precisely you can answer the questions that already exist.”

Advanced Move: Make Your Product Machine-Readable

AEO in 2026 is evolving beyond optimizing for crawlers toward declaring structured intent for AI agents. Dima Durah, Product Growth at Toloka.ai, argues the highest leverage move is to publish a machine-readable “source of truth” about your product - pricing, capabilities, and constraints - structured, explicit, and owned by you.

For example, in addition to a typical pricing page written for humans, expose a machine-readable version alongside it: a /pricing.json or /pricing.md that an agent can reliably parse.

“Agents don’t want pages, they want structured truth they can act on. The teams that win are the ones that make their context explicit, machine-readable, and impossible to misinterpret.” - Dima Durah, Product Growth, Toloka.ai

This is forward-looking but increasingly practical. If you don’t define your context for AI agents, they will infer it - and they might get it wrong.

Dima previously shared her workflow for creating an AI knowledge graph for AEO. Find the exact prompts and instructions in this GTM Strategist article.

Step 5: Build Citation Signals and Authority Off-Site

Here’s what a lot of teams miss: AI answers are heavily influenced by what other sources say about you. It’s not just about your website — it’s about your entire digital footprint.

Third-party review platforms matter enormously. G2, Capterra, and TrustRadius are among the most-cited sources when AI recommends B2B products. A complete, well-maintained G2 profile with recent reviews isn’t just for your sales team’s social proof — it’s training data for the AI that your buyers are consulting.

Other signals that feed AI recommendations:

• LinkedIn content. LinkedIn recently reported being the #1 most-cited domain for professional queries across AI search platforms. Your expert content on LinkedIn is increasingly being cited by AI systems. This is a case where employee-generated content and thought leadership have a direct, measurable impact on AEO.

• Mentions in reputable publications. Guest posts, expert quotes, original research cited by others — these are authority signals for AI, just as they were for SEO. But for AEO, they’re often more important because AI systems prioritize consensus across multiple credible sources.

• Community presence and user-generated content. Reddit threads, community forums, and Slack discussions are all sources that AI scrapes for recommendations. This is where your power users become your AEO army.

• Integration ecosystem mentions. If you’re mentioned on partner websites, integration directories, and marketplace listings, you’re building citation density that AI picks up. (This connects directly to the GTM ecosystem model — ecosystem actors generate citation signals you can’t create alone.)

José Velez shared an example of Indie Campers ($140M+ ARR): “We grew organic revenue 8.8% YoY and AI search revenue 254% in the first 5 months of working together. But to rank for competitive prompts specifically like “best RV rental company” and “RV rental Tomorrowland”, the strategy was completely different from the under-explored formats. Competitors were already covering those, so we had to create specific types of content tailored to the patterns the models were citing and secure earned mentions on the third-party sites the models pull from.”

Step 6: Monitor and Iterate (This Is Not “Set It and Forget It”)

If there’s one mistake that will waste all the work above, it’s treating AEO as a one-time project. AI outputs are probabilistic, not deterministic. They shift with model updates, competitor content changes, and citation pattern changes.

What to track:

• Share of Model: How often your brand is mentioned and recommended across AI platforms. Track monthly against competitors.

• AI-referred traffic and conversions: Set up referral tracking for chat.openai.com, perplexity.ai, and other AI sources in your analytics. Measure not just visits but conversion rate and pipeline contribution.

• Framing accuracy: Is AI describing your product accurately? If it’s wrong or outdated, you need to fix the source material. Several companies are now actively correcting misinformation that AI surfaces about their products.

• Citation sources: Which third-party sources does AI cite when mentioning you? Strengthen the ones that drive good positioning; fill gaps where competitors are getting cited and you’re not.

The companies that are winning at AEO treat it like a living system - more like product marketing than like a content calendar. They test prompts weekly, refresh key pages quarterly, and track their Share of Model the same way they’d track pipeline metrics.

The AEO Stack: Tools to Get Started

The tooling landscape is evolving fast, but here’s a practical starting point:

AI visibility monitoring:

• Profound — tracks brand mentions and citations across ChatGPT, Perplexity, Gemini; backed by $35M Series B from Sequoia

• Ahrefs Brand Radar — tracks brand visibility in AI search across major chatbots

• HubSpot Search Grader — free tool for a baseline AI visibility audit

Content optimization for AI:

• AirOps — content engineering platform helping teams see highest-impact pages and act on AI visibility data

• Surfer SEO — increasingly focused on AEO alongside traditional SEO

Manual testing (free, effective, start today):

• Run prompts yourself across ChatGPT, Perplexity, Gemini, Google AI Mode

• Track AI referral traffic in your analytics (look for chat.openai.com, perplexity.ai as referrers)

• Set up a monthly “AEO audit” cadence — 30 minutes, 15-25 prompts, document what you find

Here are also Google’s recommendations for optimizing your content for performance in AI Overviews.

AEO Is a GTM Capability, Not a Content Project

If there’s one thing I want you to take away from this article, it’s this: AEO is not a one-person content project. It’s a cross-functional GTM capability.

It touches product marketing (how your product is described and positioned), content (what you publish and how it’s structured), brand (what third parties say about you), product (what users share and how integrations are listed), and sales (the language used in decks and on calls ends up in the digital ecosystem AI scrapes).

LinkedIn recognized this when they formed a cross-functional “AI Search Taskforce” spanning SEO, PR, editorial, product marketing, and brand. They saw their B2B non-brand traffic drop up to 60% on awareness-driven topics and realized traditional SEO alone couldn’t solve it.

An important part of your AEO strategy is localization. Your buyers use AI chatbots in their natural language, so make sure that your AEO content is available and translated in the languages of your target markets.

“AEO is a compounding channel. Early wins come from listicles and competitor alternatives content, but the real compound effect shows up around months 4-6 once the models start consistently associating your brand with the category.” - José Velez, Founder & CEO, Reach

The window to build early AEO advantage is open right now. With 89% of companies not yet AI-ready, the early movers will own disproportionate share of AI recommendations in their categories. Webflow’s journey from 1.2% to 60% category share didn’t happen overnight — but it happened in months, not years.

As Dima Durah said, we’ll soon be optimizing not just for answer engines but for autonomous AI agents that browse, compare, and purchase on behalf of buyers. Companies with machine-readable, structured product truth will have a massive head start.

Here’s your minimum viable next step: Run 10 prompts today.

Open ChatGPT. Type the questions your buyers ask. See what comes back. That 30-minute exercise will tell you more about your AI visibility than any strategy deck.

Then come back here and work through the 6-step playbook. You’ve got everything you need to start.

✅ Need ready-to-use GTM assets and AI prompts? Get the 100-Step GTM Checklist with proven website templates, sales decks, landing pages, outbound sequences, LinkedIn post frameworks, email sequences, and 20+ workshops you can immediately run with your team.

📘 New to GTM? Learn fundamentals. Get my best-selling GTM Strategist book that helped 9,500+ companies to go to market with confidence - frameworks and online course included.

📈 My latest course: AI-Powered LinkedIn Growth System teaches the exact system I use to generate 7M+ impressions a year and 70% of my B2B pipeline.

🏅 Are you in charge of GTM and responsible for leading others? Grab the GTM Masterclass (6 hours of training, end-to-end GTM explained on examples, guided workshops) to get your team up and running in no time.

🤝 Want to work together? ⏩ Check out the options and let me know how we can join forces.

50% of software buying decisions involve 5 or more people - and 66% of those buyers shift priorities mid-process. Yet most ABM motions still roll everything up to a single account score and treat the entire buying committee the same way.

The result? Sales reaches out to the wrong stakeholders with irrelevant messaging. Real champions go untouched. And “orchestration” becomes just a workflow that ran.

To help you navigate this complexity, Influ2 is hosting a live panel on Tuesday, May 6 - with Devin Reed (The Reeder) and Ed VanderBush (LiveRamp). They’ll break down the 3 failure points that silently kill buyer momentum - latency, fake personalization, and false attribution - and what contact-level ABM looks like in practice.

If you’re serious about orchestrating GTM around how buyers actually buy, don’t miss this one.

Register here

Dear GTM strategist!

Claude has 10x-ed its revenue in 10 months (ChatGPT “only” 2.5x-ed it) and it took our LinkedIn by the storm.

Yet, it feels like everybody is talking about it and only a few people are harvesting massive value that Claude Code can bring to GTM teams. And it was not only a hunch - as Kyle and I are surveying our community members most people use Claude for content creation.

To do the same thing as we’ve always done more effectively (productivity). But that is just stretching the potential that Claude Code can bring to GTM - the true mastery of AI GTM comes with remaging processes and building native agentic flows.

Previously we have covered context engineering, GTM campaigns, and today I want to bring you to the weeds on how best-in-class teams such as Linear, Descript, and Canvas are building signal-driven GTM systems that actually compound over time. And we won’t stop there - you’ll get a neat GitHub repo that you can use to build your v01 of truely agentic GTM systems without relying on 20 other expensive internal tools.

So let’s go for this ride. Meet your driver Nico Druelle from The Revenue Architects and let’s do this!

When we start working with a new GTM team, we do the same thing every time: we ask to see how they use AI in their day-to-day work.

The answer is almost always the same. Someone opens a chat window. They paste in their company description, a brief ICP summary, maybe a persona or two, and then type the task. Claude produces something reasonable. They paste a slightly different version next week. The context is rebuilt from scratch every single time, then disappears when the conversation closes.

The teams we work with — at a developer infrastructure company, a design platform used by professional studios, and a half-dozen others in between — don’t work that way. Their AI has context before the session starts. It knows their ICP, their tier definitions, their signals, their competitive positioning, their current campaigns. The team doesn’t explain their GTM to Claude — Claude already knows it.

The difference isn’t the AI model or the tools. It’s that they built a repository.

What a GTM Repository Is

The concept is simpler than it sounds: a set of markdown files in a git repository that captures your GTM institutional knowledge in a structured, queryable form. Your ICP definition. Your signal library with detection methods and scoring. Your competitive battlecards. Your buyer personas with outreach hooks. Your workflow documentation.

Claude Code reads these files automatically every time it starts a session in the repo. Once the context is there, any task — research an account, score a list, build a campaign — becomes a one-line prompt.

The engineering analogy is exact: you wouldn’t rebuild your data model in a chat window every time you needed to write a query. You wouldn’t re-explain your API contracts to every new engineer. The knowledge lives in the repo. People and tools read from it.

The best GTM teams figured out that revenue knowledge should work the same way.

Why Most Teams Haven’t Built This

Because there was never a forcing function. You can paste context into a chat window and get a usable output. It works, sort of, and the alternative — actually structuring your GTM knowledge into a maintained repository — feels like overhead.

What changes the calculation is accumulation. The first time you run a signal campaign with proper context behind it versus a generic prompt, the output is noticeably better. After six months of updating your signal library with performance data, your scoring model is calibrated to actual outcomes. After a year, a new hire can open the repo on day one and understand your ICP, your signals, your competitive positioning, and your current priorities without a single onboarding call.

That compounds. Generic prompting does not.

The Architecture

A well-built GTM repository has five layers.

CLAUDE.md — The Brain

The single most important file. Claude Code reads it automatically at the start of every session. It is the persistent context layer: ICP summary, top signals, positioning, team, current priorities. Everything else in the repo is referenced from here.

The failure mode: making it too long. CLAUDE.md should be scannable in two minutes. Deep detail lives in the context files it points to.

Context — Institutional Knowledge

Six files that capture your GTM strategy in structured form.

context/profile.md — Company overview, product, deal profile by segment, and reference customers. The foundation every other file builds on. If Claude needs to describe what you do in a research brief or a sequence, it reads this first.

context/icp-definition.md — Tier definitions with explicit criteria. Not “we target Series B SaaS companies.” That is a category, not a definition. A real ICP definition includes the employee range and why, the technographic signals that indicate fit, the organizational signals (does this function exist?), explicit anti-ICP exclusions, and a qualification framework with must-haves separate from red flags.

One pattern consistent across every high-performing team we’ve seen: they maintain an ICP evolution log. Every time the ICP changes — a filter tightened, a segment removed because it’s not converting — they record what changed and why. Six months of that log is more useful than the current definition.

context/signal-library.md — Signals with detection methods, point values, decay curves, and message hooks. Not “recently funded companies.” A real signal is: “Series B announced in last 60 days, detected via Crunchbase webhook into Clay, worth 30 points, decay to 15 points after 60 days, message hook: ‘Series B is the inflection point where the ops layer either scales or becomes the bottleneck.’”

Two concepts worth understanding before you build your signal library.

Signal decay. A funding event from 150 days ago is not the same signal as one from 10 days ago. The repo scores them differently — a Series B at announcement is worth 30 points; at 90 days it’s worth 15; at 180 days it’s worth 0. This matters because most teams don’t decay their signals, which means their “active list” quietly fills with accounts that were relevant six months ago and aren’t anymore. Decay forces the list to reflect actual urgency, not historical interest.

Signal combinations. Two signals together are more predictive than either alone, and score accordingly. A Series B plus a new RevOps hire in the same week isn’t a 65-point account — it’s an 80-point account with a combination bonus. The combination tells you something neither signal tells you individually: budget exists AND someone is actively rebuilding. The repo includes a combination table with bonus scoring rules and recommended actions for each pairing.

The signal library is also where performance gets tracked. Which signals are actually generating meetings? Which ones look good in theory but produce no pipeline? A signal library that isn’t updated with results is a hypothesis document. Updated with outcomes, it is a learning system.

context/positioning.md — The source of truth for how you talk about what you do. Value pillars with proof points, a messaging matrix by persona, and a “what not to say” section that prevents Claude from generating copy that sounds good but misrepresents your actual differentiation.

context/competitor-radar.md — Battlecards that are honest about where you win and where you lose. The most useful part of a battlecard is not the feature comparison. It is the pattern: “we win when X, we lose when Y, here’s the exact language that works when they bring up [Competitor] in a call.”

context/personas/ — One file per buyer persona. Title, decision role, what they measure themselves on, how they buy, what gets their attention, what gets ignored. The sample outreach hook calibrated to each signal is what Claude draws from when building any first touch.

Skills — Repeatable Execution

A skill is a markdown file that tells Claude how to run a specific GTM task using your context. Four skills cover most of the work.

Account Research — given a domain, produce a full intelligence brief: company snapshot, stakeholder map, signal score, competitive context, and a recommended angle with a specific first-line hook. This is what goes in front of the AE before any Tier 1 conversation.

Signal to Sequence — given a signal and target segment, produce a complete campaign: trigger logic, audience definition, full sequence copy across all touches, measurement targets. Ready to load into your outbound tool. A message passes if the prospect would find it valuable even if they never buy from you. Remove the CTA. If the message still has value, it passes. If it’s pointless without the ask, it’s a pitch — rewrite it. This single filter eliminates most of what makes outbound bad.

ICP Scoring — given an account or a list, score each one against your ICP definition and assign a tier. Runs on a single account in a minute or a batch of 500 in one session.

Weekly Update — reads the current state of the repo, identifies what’s stale, and drafts every section that needs updating. Run Monday morning. Claude produces the diff; you confirm the changes. Ten minutes instead of two hours.

Workflows — The Connective Tissue

Three documents that capture how the operational processes actually work. These are not for Claude to execute — they’re decision trees and process specs for the humans running the system.

workflows/enrichment.md defines the data waterfall: which sources run first, what data quality threshold triggers the next layer, what fields are required before an account moves to active sequences. It also covers email deliverability infrastructure — domain warming, sending limits per mailbox, bounce rate thresholds — because no amount of signal intelligence survives a blacklisted domain.

workflows/signal-routing.md is the decision tree that runs when a signal fires. Customer? Suppress. Active opportunity? AE owns it. Contacted in the last 45 days? Cooldown. Still active? Score, assign tier, route to the right sequence. Without this document, signals fire and outreach happens inconsistently — some get actioned immediately, some sit for a week.

workflows/campaign-build.md maps the full campaign process from audience definition through launch to measurement review: who approves the trigger logic, who writes the copy, what QA checks run before the first email goes out, when the first review happens and what you’re looking for.

Outputs — The Archive

Every research brief, campaign brief, and sequence that comes out of the repo lives in outputs/. Six months of outputs alongside the context files that produced them is a feedback loop: you can see how your thinking evolved, which campaigns worked and which didn’t, and how your ICP has changed over time.

What This Looks Like When It’s Working

Three patterns from teams we’ve worked with, anonymized.

The signal that rewrote the ICP. A developer tooling company had been targeting “VP of Engineering” as their primary persona for 18 months — it was the intuitive choice and it had always been in the deck. After building a signal library and tracking which signals actually converted, a pattern emerged, in a weekly-update run: accounts where a “Platform Engineering” or “Developer Experience” team had been created in the last 6 months converted at 4x the rate. Not a different company type. A specific organizational moment. The ICP update that followed — adding this as a Tier 1 organizational signal — changed the shape of their outbound entirely. They found it because they were measuring.

The battlecard that got rebuilt. A security software company had lost three consecutive deals to the same competitor without being able to explain why. Their battlecard was 14 months old. After a structured loss analysis — asking the losing AEs what the competitor said, what the champion reported back, what tipped the decision — they found that the competitor had made two pricing changes and was leading with a capability they had previously deprioritized. The updated battlecard changed the next four competitive evaluations. They won two.

The new RevOps hire who didn’t need onboarding. A B2B SaaS company hired a new Head of RevOps in January. Their GTM repository had been running for 8 months. On her first day, she opened the repo: ICP definition with a log of every change made over the prior 6 months and why, signal library with performance data attached to each signal, persona files with specific outreach hooks, and a full archive of campaigns run during her predecessor’s tenure with results documented. She ran the account research skill on her first 20 priority accounts in her first week. The onboarding conversation was about strategy, not context.

None of these are dramatic wins. That is the point. The repository compounds quietly. The gains show up in ramp time, in consistency, in the patterns that only surface when you’ve been systematically tracking outcomes.

Building Yours

We built an open-source starter kit with everything described above: the templates, five skills, three workflows, two playbooks, and a fully built-out example company (Relay, a fictional workflow automation platform) with every context file populated, real sample outputs, and a campaign with actual performance data attached.

Clone it. Run the setup skill. Start from a repo that already knows your company.

The order that works:

1. Run the setup skill first — provide your domain and Claude researches your company, writes every context file, and marks what was inferred versus confirmed. The repo is 70–80% complete before you’ve answered a question. If you want to sharpen the inferred fields, the skill offers a 3-minute refinement pass after showing you the results.

2. Run the ICP Scoring skill on your current pipeline — the gaps between what you expect and what it produces tell you where your ICP definition needs work. This is the fastest way to stress-test the context Claude just built.

3. Run the Account Research skill on your top 10 accounts — use the outputs as prep for your next conversations with those accounts. See what changes.

4. After your first few campaigns, update the signal library with results — this is what turns it from a hypothesis document into a learning system.

What it will not do: Write your strategy. The ICP definition template is structure, not thinking. The signal library is only as good as your knowledge of your actual win patterns. The repo amplifies what you know. It does not replace the work of knowing it.

Keeping It Current

Setup is one afternoon. The compounding happens in the maintenance — and it’s lighter than it sounds.

Weekly (5 minutes). Update the “Current priorities” section of CLAUDE.md. This is the only file that needs to change every week. What are you focused on this week? Which campaigns are live? What’s the most important thing Claude should know before any session this week? Five minutes keeps the entire repo oriented to your current state.

After every campaign (15 minutes). Add results to the campaign output file — reply rates, meeting rates, which signals converted, which touches drove responses. This is the step most teams skip, and it’s why their signal library stays a hypothesis document instead of becoming a learning system. The signal library performance log only gets useful after you’ve run three or four campaigns and can see patterns across them.

After every win or loss (30 minutes). Update context/competitor-radar.md. What did the winning or losing AE hear? What did the competitor say? What tipped the decision? A battlecard updated in real time after a competitive deal is worth ten battlecards written in a quarterly planning session from memory.

Quarterly (1 hour). Review context/icp-definition.md and add an entry to the ICP evolution log. Did anything change? A segment that’s not converting? A new company type that’s showing up in deals? The evolution log — what changed and why, with dates — is one of the highest-value artifacts in the repo. After a year, it tells you more about your ICP than the current definition does.

The total ongoing time commitment is less than two hours a week. The question is whether that time gets spent or not. Teams that maintain the repo have a self-correcting system. Teams that don’t are running on stale context, which is almost worse than no context — it generates confident but wrong outputs.

Automating the update loop. The manual version of this workflow is: open your outbound tool, pull the week’s reply and meeting numbers, open the repo, update the signal log, update the campaign results table, run the weekly-update skill. Twenty minutes if you’re disciplined. Zero minutes if you’re not.

OpenClaw closes that gap. It’s an open-source AI agent that runs on your machine, can read and write files, browse the web autonomously, execute shell commands, and message you on Slack. The pattern: configure OpenClaw to run on a schedule, have it pull metrics from your outbound tool via browser automation, update the signal performance log and campaign results files directly, then ping you on Slack with a summary and a prompt to run the weekly-update skill. You open Slack, see “here’s what changed this week, here’s what I updated, here are the three things I need you to decide” — and respond from your phone.

The human still makes the judgment calls. Whether to change the ICP, whether a signal is underperforming enough to cut, what happened in the competitive deal on Thursday — OpenClaw can’t know those things. But the data collection, file updates, and context preparation that surround those decisions become automatic.

The /weekly-update skill included in this repo is built for exactly this workflow. Run it manually on Monday morning, or let OpenClaw trigger it on a schedule. Either way: Claude reads everything, drafts every section that changed, and asks you to fill in the parts it can’t know. Ten minutes, not two hours.

The Bigger Point

The GTM teams pulling away right now are not running more outbound, spending more on paid, or writing better copy. They have built better context. Their AI has something to work with. Their signals are calibrated to real outcomes. Their new hires don’t start from zero.

The repository is not the point. The institutional knowledge is the point. The repository is just where you put it — so it doesn’t walk out the door when someone leaves, doesn’t disappear when the Slack thread scrolls away, and doesn’t have to be rebuilt every time you run a new campaign.

Build it once. Update it when things change. Let it compound.

Get the Starter Kit

GitHub: github.com/KarlRaf/gtm-starter-kit

What’s included:

• CLAUDE.md template (pre-structured, not a blank form)

• 6 context templates: company profile, ICP definition, signal library, positioning, competitor radar, persona

• 5 skills: Setup (auto-populate from domain), Account Research, Signal to Sequence, ICP Scoring, Weekly Update

• 3 workflows: Enrichment (with deliverability), Signal Routing, Campaign Build

• 2 playbooks: New Signal Response, Competitor Switch

• sync/ folder with starter scripts for pulling campaign data into the repo automatically

• Complete example company (Relay) — every context file populated, two sample outputs, a live campaign brief with sequences and performance tracking

If you enjoyed this article, you can explore more about Claude Code and context engineering for GTM in these recent articles:

• How to Build GTM Campaigns Your Prospects Would Pay to Receive

• The GTM Guide to AI Context Engineering

✅ Need ready-to-use GTM assets and AI prompts? Get the 100-Step GTM Checklist with proven website templates, sales decks, landing pages, outbound sequences, LinkedIn post frameworks, email sequences, and 20+ workshops you can immediately run with your team.

📘 New to GTM? Learn fundamentals. Get my best-selling GTM Strategist book that helped 9,500+ companies to go to market with confidence - frameworks and online course included.

📈 My latest course: AI-Powered LinkedIn Growth System teaches the exact system I use to generate 7M+ impressions a year and 70% of my B2B pipeline.

🏅 Are you in charge of GTM and responsible for leading others? Grab the GTM Masterclass (6 hours of training, end-to-end GTM explained on examples, guided workshops) to get your team up and running in no time.

🤝 Want to work together? ⏩ Check out the options and let me know how we can join forces.

The GTM10 Awards recognize the best go-to-market leaders in software across Growth, Marketing, Sales, Customer, and Partnerships. Powered by Cello and Lightspeed Venture Partners, with a jury featuring leaders from Zapier, Vercel, Airtable, Sequoia, a16z, and Stripe.

I’m glad to be on the jury this year for the Growth category.

50 winners get announced starting June 1 and featured on the iconic NASDAQ billboard in NYC.

Know a GTM leader who deserves recognition? Nominate them. Think you qualify? Claim your own nomination.

Nominations close May 5.

Nominate now

Dear GTM Strategist,

You know those outbound messages that make you want to throw your laptop out the window?

“Congrats on your new role!”

“I noticed your company is growing.”

“Company X deserves more SEO backlinks.”

- followed by a pitch for something you’ll never buy from someone who clearly knows nothing about you.

Here’s why it won’t end anytime soon: mass outbound “spray and pray” tools - now powered by AI - generously include such templates, and sales noobs and lazy sellers eat them up.

But if you send out 500 emails and book zero meetings, treat that as a sign that something needs to change.

Claude Code and Jordan Crawford to the rescue. Jordan is the founder of Blueprint GTM, where he helps teams learn and work with Claude Code, and a co-creator of Cannonball GTM - a Substack that I cannot recommend enough!

And he’s here to help

Instead of sending out 500 “congrats on your recent promotion to the VP of strategy at X…”, you will learn how to craft an email that intelligent life will actually reply to because it talks to their pain points and adds value.

So in today’s post, I want to share what we learned and share more tips on how you can apply it to build better campaigns.

What you’ll learn:

• Context engineering - how to load AI with deep, specific knowledge about your ICP, market, and to find data sources to get 10x leverage.

• How to set up Claude Code as your GTM operating system with compounding context, persistent memory, and the ability to query databases, write scripts, and deploy sub-agents.

• Why permissionless value prop (PVP) - a message that is independently valuable to the recipient, whether they buy from you or not - beats cringe personalization.

Why Most AI-Generated Outreach Fails

Problem: Everybody has access to the exact same signals and tools - and those are not the best.

Job changes, funding rounds, LinkedIn activity, technographics - you can get all of this from a dozen enrichment tools. And so can every one of your competitors.

When everyone enriches from the same signal pool, the “personalization” is identical across the board. You’re all writing the same email to the same person with the same data points.

Solution: Take a step back - deep dive into your ICP, buying committee, and fall in love with their problems

The generic enrichment layer has become the baseline. The teams pulling ahead are the ones who layer proprietary context on top: their own win/loss patterns, domain-specific data sources, industry signals that no enrichment vendor packages out of the box.

As you build your proprietary intelligence, you unlock first degree signals.

Jordan Crawford put it perfectly: “We are currently limited not by our tools anymore, but by our imagination. What’s the greatest thing you can imagine with the best possible data?”

Most teams use AI to write faster. Very few use it to think deeper about what would actually be valuable to say.

The question worth asking: what is the most valuable thing I could say to this person, and what data would I need to say it?

Often in GTM, you have one shot to do it - so you better do it right

That’s a fundamentally different starting point - and it leads to fundamentally different messages. And it motivates you to put in more thought and love instead of thinking that changing names and companies is the level of personalization you need.

Context Engineering: The Skill That Separates 10x From 1x

Context engineering goes way beyond prompt engineering. A clever prompt still works with whatever the model already knows. Context engineering is about loading an AI system with enough specific knowledge - ICP docs, call recordings, competitive intel, product data, market signals, database access - that it stops guessing and starts reasoning like your best rep.

Why Claude Code specifically? Because, unlike a regular Claude or ChatGPT chat, Claude Code offers compounding context:

• Persistent memory. It remembers what you’re working on across sessions. Each client gets a folder with accumulated knowledge that builds over time.

• Skills. These MD files are reusable workflow instructions - think of them as playbooks that teach Claude Code how to perform a specific set of tasks - like validating a LinkedIn profile, querying a specific database, or structuring a particular type of analysis. (Here’s Anthropic’s official guide for building Skills).

• Sub-agents. Claude Code can spawn specialized agents for subtasks without polluting the main context. Think of it like a manager delegating to employees - the manager doesn’t need to know every detail of how the floor got swept, just that it’s done.

• Script execution and API calls. It can write and run actual code, query databases, call APIs and write to other systems, and save results locally. Think of it as an operating system, not a chatbot.

The first campaign is the heavy lift. Every one after that builds on the context you’ve already loaded - so the intelligence compounds while the time to deploy drops.

The setup on your computer should be simple: one folder per project or client. Inside that folder lives everything Claude Code needs - the company context, the ICP definition, access to relevant data sources. When you open that folder, Claude Code already knows what it’s working on and it doesn’t mix it with the context and memory from other projects.

“The process matters more than the prompts. Long prompts are the past. What matters is your goal, the order of operations, and the context you load before you start building.”

Example: A Campaign for GAF That You’d Actually Send

To demonstrate what building a GTM campaign with Claude Code looks like, Jordan picked a deliberately unexpected company: GAF, one of the largest roofing manufacturers in the world. GAF sells roofing tile through independent contractors. About as far from the B2B SaaS bubble as you can get.

Why? Because great outbound isn’t limited to B2B SaaS. If you can find discoverable pain and public data, you can build a killer campaign for anyone.

Here’s what happened in roughly 30 minutes:

Step 1: Context loading. Jordan told Claude Code to research GAF’s website and understand the business - who their customers are, how they tier their contractors, what their reward programs look like. No manual briefing doc. Just: “Go understand this company.”

Claude Code ran web searches, built a markdown (MD) file summarizing GAF’s business model, contractor ecosystem, and competitive landscape - and saved it as persistent context.

Because context windows in AI tools are limited (and can fill fast with data-intensive research), storing relevant data in MD files ensures that Claude can always reference key information while you clear the context window for a new task. You can even ask Claude to save important info to its memory.

Step 2: The plan. This is where it got interesting. Think of where you could find relevant, unique data that is not sourced in standard enrichment tools and can give your messages an edge - like public databases or industry associations. For the GAF case, Jordan used the Shovels database, which contains 185+ million building permits. He told Claude Code: figure out what intelligence in this data would be incredibly valuable to GAF.

That’s the magic of Claude: you can tell it your goal and it will find the best way to do it.

Claude Code found 19.6 million roofing permits in the Shovels database, with 276,560 permits mentioning GAF. Based on this, it identified the play on its own: upsell targets. Contractors who already use some GAF products but give 60–90% of their business to competitors like Owens Corning and CertainTeed. Claude Code even identified GAF’s competitors from the permit data without being told who they were.

Still, use your brain. You should be the final judge of the plan and ruthlessly tell Claude where it went wrong. In our example, the pitch it initially suggested was wrong and Jordan told it how to correct it.

Step 3: Data queries and message generation. Claude Code queried the database multiple ways - loyalty splits, geographic white space, competitive displacement targets, brand share trends over time. It found 154 high-quality contractors that represent massive upsell opportunities.

Then it identified the right decision-makers and drafted messages with actual data:

“There are 200+ roofing contractors who’ve used GAF in at least 10% of their permitted jobs, but they’re giving 6y just haven’t committed. Here are the top five - with their revenue, Google reviews, total permits, and exact brand mix.”

That’s market intelligence disguised as outreach. And it compels a response because it’s independently valuable - the recipient benefits from reading it whether they take a meeting or not.

Jordan calls this a Permissionless Value Prop (PVP) - a message so valuable that prospects would literally pay to receive it. You’re giving away an insight, and the meeting is just a natural next step. For this, you don’t need a large database of potential leads - but a highly focused one.

“I’m giving you valuable data on your market. That’s worth a response. Nobody ever called market intelligence ‘creepy.’”

👉 Watch the full livestream recording to see how Jordan Crawford built this whole campaign live.

The Playbook: How to Build Your Own Context-Engineered Campaigns

Ready to try this yourself? Here’s the framework Jordan uses with every client.

1. Invert from your best deals

This is Charlie Munger’s advice applied to GTM. Download all your call recordings (or ask Claude Code to create a script to get them for you). Deploy Claude Code against every transcript and ask: what likely made this person convert? Compare closed-won against closed-lost.

Then ask the harder question: what public data could have predicted this? For each conversion signal, search for publicly discoverable evidence. If your best customers tend to be companies with outdated tech stacks going through rapid hiring - can you find that signal before they ever hit your website?

2. Find discoverable pain

Not every company is a good fit for this approach. You need two things: a differentiated product and high public data availability. Regulated industries - healthcare, construction, finance - have a treasure trove of public filings, permits, compliance data, and regulatory signals.

But even in lower-data environments, you can often find something - job postings, technology stack signals, press releases, SEC filings.

3. Build the context folder

This is the “context engineering” part. Before Claude Code writes a single message, load it with everything your best rep knows:

• ICP documentation

• Call transcripts and win/loss analysis

• Competitive intelligence

• Product positioning and objection handling

• Access to relevant data sources (Exa.ai, Blitz API, OpenWeb Ninja, Apify, or whatever fits your market)

4. Plan before you execute

Jordan uses Claude Code’s plan mode religiously. This will get Claude to draft the strategy right before burning tokens on heavy lifting. What data do you need? How should you query it? What’s the angle? The plan will be saved as an MD file, so Claude can reference it even as its memory resets across tasks - so it has to be airtight.

5. Go deep, not wide

Jordan’s “drill, not peanut butter” analogy is the best mental model here. Don’t spread thin across thousands of accounts. Pick a specific segment where you have deep data and a strong value prop. Build one killer campaign. Prove it works. Then expand.

“Scaling this to a thousand companies without doing the hard thinking first is like pointing a thousand cars off a cliff. The best campaigns start narrow and go deep.”

But I’m Not Technical - Can I Still Do This?

If you can write a Google Doc, you can use Claude Code.

The terminal might seem scary at first glance, but you will quickly get used to it. There are just a few commands you actually need:

• cd + drag and drop a folder (to navigate to your project)

• claude (to start Claude Code)

• shift + tab (to switch to plan mode)

• clear (to reset the context window)

• ctrl + V (to paste - note that it differs from cmd + V even on macOS)

That’s it. Everything else, you just write (or even talk - speech-to-text tools like Wispr Flow can be a huge time-saver). You describe what you want to do. Claude Code figures out how. If you don’t know how to do anything in the terminal, just ask Claude in chat and it will guide you through every step.

The Bigger Picture

Any SDR tool can send a thousand emails in an hour. The real unlock is deep research that is scalable.

The analysis Jordan did live - querying building permits, identifying competitor usage patterns, finding geographic white space, building data-rich messages for specific decision-makers - would have taken a research team weeks. Claude Code did it in 30 minutes.

The teams that win in 2026 won’t be the ones sending the most emails. They’ll be the ones sending messages that make prospects think: “How do they know this about my market?”

That’s the bar now. Context engineering gets you there.

Here are your next steps:

• Open Claude Code this week and start with one thing: load it with context about your ICP. Upload your best call transcripts. Ask it to identify what made your best customers convert. Work backward from there.

• Grab my 12 Claude Code skills for GTM. I turned my entire GTM methodology, tested with 1,000+ companies, into a set of Claude Code skills that cover everything from competitor analysis and positioning to pricing, launch assets, outbound, and ABM. 100+ tasks across 12 phases, where each phase builds on the last. Add them to Claude Code or Cowork, ask for what you need, and the right skill activates with tailored output based on your product context. Get them here for free on GitHub.

The AI model is table stakes - everyone has access to the same LLMs. The context you feed it is your moat. Start building it.

Claude Has Taken Over GTM: Now Tell Us How You’re Actually Using It

Q1 2026 made one thing crystal clear: Claude didn’t just enter the GTM stack. It started eating it.

Teams are using it to scrape niche directories for non-obvious prospect lists, build trigger-based outbound sequences, create programmatic AEO pages at scale, reverse-engineer competitor ads, build custom lead scoring logic, and clean up CRMs without touching a single line of code. And that’s just what we already know about.

Kyle Poyar (Growth Unhinged, 80K subscribers) and I are running the first-ever Claude Code Pulse Survey to map what GTM teams are actually building with Claude right now, in production, not what looks good in a YouTube demo.

This is what you get for 5 minutes of your time:

• Early access to the full results before anyone else, including real workflows and setups from practitioners across outbound, content, CRM, growth, and sales

• A chance to be featured as a guest contributor in Growth Unhinged or GTM Strategist. If your setup is interesting, we want to write about it.

We’re going all-in on Claude use cases and workflows in Q2, and we need your help figuring out what’s already working (and what’s still broken) so we can curate content that brings tangible value to your business.

👉 Take the survey

It takes under 5 minutes. I did it in 3 minutes. Can you beat me?

✅ Need ready-to-use GTM assets and AI prompts? Get the 100-Step GTM Checklist with proven website templates, sales decks, landing pages, outbound sequences, LinkedIn post frameworks, email sequences, and 20+ workshops you can immediately run with your team.

📘 New to GTM? Learn fundamentals. Get my best-selling GTM Strategist book that helped 9,500+ companies to go to market with confidence - frameworks and online course included.

📈 My latest course: AI-Powered LinkedIn Growth System teaches the exact system I use to generate 7M+ impressions a year and 70% of my B2B pipeline.

🏅 Are you in charge of GTM and responsible for leading others? Grab the GTM Masterclass (6 hours of training, end-to-end GTM explained on examples, guided workshops) to get your team up and running in no time.

🤝 Want to work together? ⏩ Check out the options and let me know how we can join forces.

This post is supported by Schematic — monetization infrastructure for product & engineering teams.

Schematic decouples billing from code, giving product teams direct control over pricing, packaging, limits, add-ons, credits, trials, migrations, and upgrades — without burdening engineering.

The result: faster pricing experiments, better conversion, more upgrades, and less revenue leakage. Works natively with Stripe Billing. Used by high-growth teams like Plotly, Pagos, TightKnit, and Zep.

Try Schematic free

Dear GTM Strategist,

Everyone talks about how AI will make GTM cheaper because you can just vibe code your own tool and cancel expensive subscriptions. But an interesting trend is happening instead.

AI tools are becoming more expensive to run, not less. And the reason has very little to do with pricing decisions. It has to do with how AI software actually works.

Clay’s recent pricing change is a good example.

When they announced the update, reactions were all over the place. Some users celebrated the fact that data costs dropped significantly. Others quickly calculated that the same workflows would now cost much more.

Both reactions were technically true.

Clay did not simply change prices. They changed how costs are exposed inside the product.

For the past two decades, SaaS followed a predictable model. Companies sold access to software, usually priced per seat, while the cost of serving each additional user was close to zero.

AI tools break that model. Every AI-powered workflow consumes something real behind the scenes. Tokens. API calls. Model inference. Compute.

As usage grows, so do the underlying costs. Which means the economics of AI products look very different from traditional SaaS.

Clay’s pricing change simply makes that reality visible. And it highlights a much bigger shift happening across the AI software landscape.

In this post, we will look at:

• Why AI tools are forcing new pricing models

• Why credits and usage-based pricing are spreading across SaaS

• How teams are tackling variable costs of AI tools

• and what this means for the future economics of AI-powered GTM stacks

Let’s get into it.

Clay’s Pricing Change Is a Signal

Clay introduced a new pricing mode last weekl. At first glance, the announcement sounded simple. Data costs inside the platform dropped significantly. In many cases, the reduction was between 50 and 90 percent.

That part was easy to understand. Data enrichment has always been one of the most expensive parts of running Clay workflows, so cheaper data was welcomed by many users.

But the rest of the update revealed something more interesting.

Clay separated the costs of two things that used to be bundled together:

• Data credits

• Workflow actions (platform usage)

Data credits now pay for enrichment and external sources. Actions pay for running workflows and executing tasks inside the platform.

In other words, Clay is no longer charging only for access to the product. It is charging for the work the product performs. As Kyle Poyar pointed out in his analysis, this pricing now splits out the value/platform piece (high margin) vs. AI tokens (low margin).

The rollout itself was a masterclass in comms: coordinated internal/external messaging, CEO posts, video walkthroughs, and all existing customers grandfathered on the legacy model.

It created two very different reactions in the community.

Some users calculated that their workflows would now be cheaper because enrichment costs dropped so much. Others realized that heavy automation would now generate higher bills.

Both interpretations were correct. It also depends on how they are using Clay.

The pricing model did not simply go up or down. It exposed where the real costs actually sit. And those costs are not primarily inside the Clay product itself. They come from everything happening underneath the interface.

Every automated workflow triggers multiple external services. Data providers charge per record. AI models charge per token. Infrastructure charges for compute and storage.

The more work the workflow performs, the more resources it consumes. Traditional SaaS products could hide those costs because they were small and predictable.

AI products cannot.

Clay’s pricing update simply makes those economics visible.

AI Products Cannot Be Priced Like Traditional SaaS

To understand why Clay made this change, it helps to zoom out. For most of the past two decades, SaaS companies followed a predictable model.

They sold access to a product, usually priced per seat or per account. Once the software was built, the cost of serving each additional user was relatively small.

That is why SaaS margins became so attractive. Revenue scaled faster than costs and ARR (Annual Recurring Revenue) was the name of the game.

AI products behave differently. Every time a user runs an AI workflow, the workflow consumes AI tokens. Those costs do not disappear as usage grows. In many cases they grow directly with usage.

This is one reason pricing teams have become unusually active recently. According to the PricingSaaS Trends Report, pricing changes across SaaS companies increased roughly 15 percent year over year, while packaging changes increased more than 20 percent as companies experimented with new pricing structures.

One trend stands out in particular: credit-based pricing models grew more than 120 percent.

Credits allow companies to connect price more closely to consumption. Instead of charging only for access to the product, they charge for the work the product performs.

This shift is already visible across many AI products. Figma introduced AI credits for certain features. Data platforms charge per request or per record processed. Some AI assistants are priced per resolution or per automated task.

Kyle Poyar recently described this shift as moving from selling software access to selling completed work. He also analyzed how

The pricing unit is no longer a seat. It is the task the system performs. And this is reflected in AI-based pricing models.

The Hidden Cost of the AI GTM Stack

Once you start looking at AI tools through the lens of usage, something else becomes visible. Most GTM teams are no longer buying a single product.

They are assembling a stack of services that work together.

A typical outbound workflow today might look simple on the surface.

• A contact gets enriched.

• The company is analyzed.

• A personalized message is generated.

But behind the scenes, that workflow may trigger several systems.

An LLM analyzes the company or contact. Another model generates the message. Automation tools move the data between systems, and users can “bring their own keys” - connect the their preferred tools with API keys to source the data directly.

Each of those steps has a cost attached to it.

• Data providers charge per record.

• AI models charge per token.

• Infrastructure charges for compute.

None of those costs is visible when the workflow is first built. They only appear once the system starts running at scale. This is why many teams are beginning to notice their AI budgets growing faster than expected. As 2026 SaaS Management Index shows, 61% of IT leaders have already cut projects due to unplanned SaaS cost increases.

A workflow that feels inexpensive during testing can become costly when it runs thousands of times per week. At that point the GTM stack starts to behave less like traditional software and more like infrastructure.

And infrastructure is almost never priced per seat.

Pricing Models Are Starting to Reflect This Reality

Once you understand the economics behind AI products, the recent wave of pricing changes makes sense. Companies are searching for ways to connect pricing to the actual work their systems perform.

Several models are emerging.

Some companies introduce monthly credit allowances that reset each billing cycle. Others charge once usage exceeds predefined limits.

Hybrid models are also becoming common. A base subscription provides access to the product, while usage-based components capture the cost of AI features.

A few companies are going even further and pricing based on outcomes.

Support platforms charge for resolved conversations. Automation platforms charge for executed workflows. AI assistants charge for generated outputs. All of these approaches attempt to answer the same question.

What is the most honest unit of value for AI software?

The industry has not settled on a single answer yet - ideally completed outcomes, but that’s not always feasible to measure. However, one thing is becoming clear. Seat-based pricing alone will not capture the economics of AI systems where usage directly drives cost.

How Top GTM Teams Are Tackling AI Costs

I asked world-class companies in my network what the variable and sometimes unpredictable AI costs mean for them and how they are tackling this.

Amos-Bar Joseph, CEO and co-founder of Swan AI, famously shared his $50K monthly Anthropic bill.

However, a month later, he shared how he optimized the costs - slashed the bill to $27K while doubling AI usage (Swan grew from 200 to 400 customers that month).

What changed? Swan’s AI agent had been operating through a translation layer - converting every action into human-language instructions that then had to be decoded into code. By removing that layer and rebuilding the agent to operate directly in code - calling APIs and databases natively - they cut costs by nearly half while doubling usage.

The lesson from Amos: the biggest cost savings don’t come from prompt optimization or model switching, but from rethinking the AI architecture itself.

Kareem El Muslemany (senior monetization manager at Zoom and author of Pricing Code) highlighted that AI is novel, and costs can escalate unpredictably - 80% of companies in a recent survey were 25% or more over their AI budget. Customers can use cost and ROI calculators from vendors to budget for AI costs. Variable costs can be optimized by using lower-cost services or AI models when they suffice, using internal data/tools, and having fewer steps in workflows. As a final check, spend caps can be implemented to avoid unplanned overages.

But as Kareem also points out, pricing leaders should be mindful of pure-token costs. In his post, he describes that for pure coders using Claude’s API alone, raw token costs don’t always capture the true value. Code Review, one of its highest-value developer workflows so far, is billed on token usage at $15–25 per review. That logic made sense when Claude was mainly infrastructure, but Code Review is valuable IP. Monetizing on credits alone undersells its incremental value.

Fynn Glover, co-founder and CEO of Schematic, said that how you price against rising AI costs is a strategy question. “It depends entirely on two things: your capitalization and your pricing philosophy based on your stage and your capitalization.”

If you’re well-funded and in a land grab, your philosophy might rightly be, “for now, let’s not let input costs drive our pricing.”

If you’re bootstrapped or undercapitalized and focused on demonstrating pricing power, then yes, cost discipline matters a lot.

“The mistake I see most often: companies defaulting to cost-plus pricing because their AI costs are rising. I think classic pricing theory is quite instructive here. The moment you anchor your price to your costs instead of your customer’s value, you’ve told the market exactly how to think about you and it’s not favorably. Cost-plus can work for pass-through add-ons. For your core product? It erodes perceived value and gives up pricing power,” said Fynn.

The New Skill GTM Teams Will Need

This shift creates a new challenge for GTM teams. The efficiency of AI workflows suddenly matters.

In the past, teams rarely needed to think about the internal mechanics of the tools they used. You bought the software, added users, and focused on adoption. With AI systems, the structure of the workflow itself becomes important.

A poorly designed workflow may trigger unnecessary enrichment calls. Prompts may be longer than needed. Automation loops may multiply the number of model calls.

None of this is visible in the interface. But it shows up very clearly in usage-based pricing.

This is why some teams are starting to treat AI workflows almost like engineering systems. They monitor how often processes run. They optimize prompts to reduce token usage. They simplify automation chains to avoid unnecessary calls.

The goal is not only to build workflows that work. The goal is to build workflows that work efficiently.

That is a new skill set for most GTM teams.

It sits somewhere between operations, data, and product thinking. Teams that learn to design efficient workflows will get more output from the same tools. And over time, that advantage compounds.

AI software is changing the economics of the SaaS model. The future GTM stack will not only be evaluated based on features.

It will also be evaluated based on how efficiently it converts compute into outcomes.

The teams that understand those economics early will build systems that scale without surprises.

To hear directly from top pricing experts how they’re tackling AI monetization in 2026, join me on March 26 for a livestream with Kareem El Muslemany (Monetization at Zoom), James D. Wilton (Monevate, ex-McKinsey, author of Capturing Value), Jasdeep Garcha (Product & Monetization at Vercel), and Fynn Glover (CEO, Schematic).

We’ll dig into credits vs. usage, predictability vs. alignment, outcome-based vs. output-based pricing, and the growing gap between the value your AI product creates and the revenue you actually capture. No slides, no pitches - just what’s working and what’s not.

Join here

✅ Need ready-to-use GTM assets and AI prompts? Get the 100-Step GTM Checklist with proven website templates, sales decks, landing pages, outbound sequences, LinkedIn post frameworks, email sequences, and 20+ workshops you can immediately run with your team.

📘 New to GTM? Learn fundamentals. Get my best-selling GTM Strategist book that helped 9,500+ companies to go to market with confidence - frameworks and online course included.

📈 My latest course: AI-Powered LinkedIn Growth System teaches the exact system I use to generate 7M+ impressions a year and 70% of my B2B pipeline.

🏅 Are you in charge of GTM and responsible for leading others? Grab the GTM Masterclass (6 hours of training, end-to-end GTM explained on examples, guided workshops) to get your team up and running in no time.

🤝 Want to work together? ⏩ Check out the options and let me know how we can join forces.

This newsletter is kindly supported by Attio - the AI-native CRM built for modern GTM teams. If you’ve ever battled a rigid legacy CRM, Attio fixes that with clean data, real-time visibility, and workflows that actually speed you up.

Today, 5,000+ fast-moving companies rely on Attio’s flexible data model, modern design, and builder-friendly setup to shape their CRM around how they work - not the other way around. It’s built for founders, revenue ops, and GTM teams who need a system that adapts as fast as we do.

So if 2026 is the year you finally break up with your legacy CRM or start fresh on a new project, this is the perfect moment to take Attio for a spin.

Try Attio for free

Dear GTM Strategist!

Before we jump into another exciting round of AI playbooks, we need to talk about the infrastructure we are operating in. Most conversations about AI in go-to-market focus on tools and tactics. How to automate outreach. How to build agents. How to scale workflows.

These conversations are exciting, but they often miss the bigger issue.

Artificial intelligence did not fundamentally change GTM.

It exposed how many of our systems were already broken.

For years we built revenue engines that relied heavily on human effort. People writing cold emails. People updating CRM records. People manually interpreting customer feedback. It worked when capital was cheap and headcount was easy to justify.

Those systems are now under immense pressure.

Jacco van der Kooij, the brilliant GTM strategist leading the Winning by Design movement, frames the shift in a very intriguing way.

“For years, we talked about keeping the human in the loop. Today the real challenge is to close the loop.”

Closing the loop means building systems where learning happens continuously. Insights move quickly through the organization. Decisions are based on structured data rather than fragmented opinions.

That requires something many GTM teams would rather avoid. It requires looking at the infrastructure behind how revenue actually works.

After reading Jacco’s chapter in Momentum’s Ignite Your GTM with AI, I spent months trying to get him to talk about this topic with us. If AI is going to work inside GTM systems, we first need to understand the architecture those systems run on.

And that architecture is often messy.

Teams are building AI workflows on top of fragmented data, disconnected tools, and processes that were never designed to scale. Instead of creating leverage, this often creates more complexity.

This conversation is about fixing that foundation.

It is not a light read. But if you are responsible for leading GTM teams in the AI era, it will change how you think about your systems.

In this edition, we’ll cover:

• Why three “black swan” events colliding at once created the perfect storm for GTM

• How AI-native companies build compounding revenue machines instead of linear sales motions

• The four data crises you must fix before AI can actually help you

• Why the excavator analogy should change how you think about your team

• How to optimize for buying instead of selling, and why that distinction matters more than ever

Let’s get into it.

The Macro Shift and the AI Revenue Machine

Artificial intelligence did not change go-to-market overnight. It simply exposed what was already broken. And this massive shift did not happen in a vacuum. Jacco points out that it is the direct result of what he calls the Three Black Swan Convergence - three compounding macro events colliding at the exact same time.

COVID-19 fundamentally changed how we work and interact. The sudden end of zero interest rates completely changed how we fund businesses. And artificial intelligence changed how we build and scale products. Each of these on its own would be disruptive. Together, they exposed a massive systemic flaw in how software companies operate: for the last decade, we over-indexed on cheap human labor instead of building truly intelligent systems.

When money was practically free, founders could raise millions on a mediocre idea and hire armies of salespeople to brute-force their way to revenue. We bloated our recurring revenue models with manual, unstructured work. Humans writing cold emails. Humans manually updating CRM records and parsing through customer feedback. Every single time you insert a human into a repetitive data process, you introduce delay, bias, and error. For years we compensated for these slow feedback loops with more effort and more headcount. The system was highly inefficient, but it was still moving. Now that inefficiency is completely visible.

Jacco explained this perfectly:

“The main change that we see over the last two years is that AI has revealed how recurring revenue is supposed to work. And it did that by removing the influence of people in an unstructured way, that it actually slows the system down.”

AI-native companies operate entirely differently. They strip away unstructured human influence and build closed-loop systems. If a new user segment starts converting faster today, the system detects it immediately and automatically routes more resources to that specific segment. Learning takes minutes or hours, not months.

This speed is driven by what I’d call User-Decider Convergence. The old software model relied on top-down procurement - you sold to the executive, and they forced their team to use the tool over a nine-month sales cycle. AI natives flip this entirely. The person using the product is also the champion and the decider. They experience the pain, find the solution, and feel the impact instantly. They swipe their own credit card. Adoption spreads organically from the bottom up, and then they upgrade their own account and invite their entire team to join.

This is how you build a compounding revenue machine. This is how a company scales from $1M to $100M without hiring a thousand salespeople. You don’t just buy a new AI tool - you change your entire operational architecture.

The Four Data Crises

Before AI can act as your growth engine, you must fix your data infrastructure. You cannot scale a broken system, and if your underlying data is a mess, adding artificial intelligence will only help you make expensive mistakes much faster. Most traditional software companies are currently paralyzed by their own tech stacks, trying to deploy advanced AI models on top of a broken foundation.

Here are the four critical data crises you have to solve first:

Unstructured Data. Ninety-five percent of your customer conversations remain unformatted and completely unused. Your sales team has hundreds of calls every single week, and they take highly subjective notes based on gut feelings. This unstructured data is completely useless to a machine - AI cannot parse a vibe. You must use a strict methodology like the SPICED framework to turn human conversations into structured, machine-readable data.

Incompatible Data. Your tools do not speak the same language. Sales call recordings live in one platform. Pipeline revenue lives in your CRM. Customer support tickets are stuck in a completely different system. They exist in isolated silos, and you cannot build a unified growth engine if your data is fragmented across six different dashboards that refuse to integrate.

Disconnected Data. This is exactly what breaks the compounding growth loop. A customer success manager might learn exactly why a major client churned on a Tuesday, but that critical insight never flows back to the marketing team running the acquisition campaigns. The system is entirely disconnected - the right hand has no idea what the left hand is learning.

Incomplete Data. Gaps in your data capture are incredibly dangerous in the AI era. If you feed an AI model incomplete data, it will not just misfire - it will actively mislead you. It will confidently tell you to target the exact wrong segment and burn your entire marketing budget in the process.

To fix these four crises, you need to build a real-time data infrastructure. This is the core difference between the old way of operating and the new AI-native approach: you need a closed-loop system where data flows continuously. Learning from a customer cannot take an entire quarter. It must take hours.

Here is how Jacco van der Kooij explained this urgency:

“What do I mean with real time? Well, look, if you learn something happens in a region or if you learn a particular conversion rate is in a particular state and it moves down. It cannot take months before that propagates and leads to action. It is got to take minutes, hours.”

If a conversion rate drops abruptly, your system needs to adjust that same afternoon. If a new and highly profitable buyer persona suddenly appears in your funnel, you need to throw more resources at that fire immediately. You can no longer afford to wait for a monthly reporting meeting to figure out what the market wants.

The Excavator Analogy and the 30/70 Talent Split

We are currently forcing our most expensive talent to do manual labor. We pay human beings to write generic outreach messages, scrape lead lists, and update pipeline stages. This makes absolutely no sense.

Jacco uses a brilliant construction analogy to explain this shift. If you need to dig a massive hole, you can hire four people with shovels - they will eventually get the job done. Or you can bring in an excavator that digs faster, deeper, and with far more accuracy. AI is the excavator. But here’s the important nuance: when you bring the excavator onto the site, the humans don’t just disappear. Their roles completely change. You suddenly need an inspector to make sure the machine doesn’t hit a gas line. You need an architect to verify the foundation is square.

This is how Jacco explains it:

“The human beings are taking away the role of physical labor, the shovel, and are now becoming humans that oversee if the work is done properly.”

Your team should not be writing the spam. They should be managing the AI agent that generates the campaign - reviewing the target list, tweaking the context of the messaging, approving the final output. Once they perfect that process, they build an AI agent to check the first AI agent. This is a fundamentally different way of working, and it requires fundamentally different skills.

This shift brings us to a harsh reality about talent. AI will not fix your mediocre performers. Most leaders assume that AI will magically elevate the bottom 70% of their team - that it will turn struggling reps into top performers. It will not. If a rep doesn’t understand the buyer’s pain points, giving them an AI tool just helps them send irrelevant messages much faster.

The real magic happens at the top. AI gives your top 30% superpowers. These are your relationship builders, the complex problem solvers who actually understand the market. When you remove their manual data entry and administrative tasks, they operate at unprecedented levels. So stop trying to use AI to save the bottom 70%. Instead, architect systems that eliminate the need for routine execution work entirely.

Avoiding the Sales Tech Sidestep

Sales technology took a massive wrong turn over the last decade. The earliest tools were built to help us use data to work smarter - they were optimized for intelligence. Then we took a detour. Later tools completely abandoned intelligence and optimized for pure volume instead. We built massive machines to automate cold outreach and scale email spam. AI is now forcing us to return to that original principle: using data to reach the right people at the right time with genuinely valuable information.

This is where early-stage founders have an unfair advantage. If your company is under $5M in revenue, you are in the best possible position. You don’t have to unlearn bad habits. You don’t have a massive legacy tech stack to dismantle. You can build a clean AI-native system from scratch.

Jacco shared the exact tactical workflow with me, and it is incredibly simple:

1. Use LinkedIn Sales Navigator. Set strict filters to pull a highly targeted list of 200 prospects.

2. Export that list and feed the profiles into an AI model like Claude. Ask the AI to analyze those 200 people and find the exact common problem they all share right now.

3. Use a tool like HeyReach to build a highly contextual campaign based on that specific research.

A few years ago, this level of research and execution would take a team of human SDRs several days to complete. Today, a single founder can do the entire thing in three hours. You don’t need a massive team to compete anymore - you just need a smarter architecture.

Optimize for Buying

We need to talk about buyer psychology, because this is where everything connects. People absolutely hate being sold to - but they love to buy. If you stop someone on the street to aggressively sell them a timeshare, they will run away. If you put that same person on Amazon, they will happily spend hours filling their cart. The motivation is completely different when the buyer is in control.

“Most people hate being sold but they love to go buy. And that means that the sales process should not be based around selling and optimizing selling; it should be based on buying and optimizing buying.”

The old software model completely misunderstood this dynamic. We built massive, complex processes optimized entirely for the seller - mandatory discovery calls, artificial demo gates, aggressive outbound follow-ups. We tried to control the timeline. The new AI-native model uses technology to optimize the buying process instead. It removes friction. It uses AI to figure out exactly what specific research or proof the customer needs to make a purchasing decision today. It doesn’t blast them with templates - it finds the patterns in their problems and offers the exact solution they are actively looking for.

This is where true scale happens. When you optimize for buying, you activate compound growth loops. You achieve exponential growth when your current user essentially sells your product to the next prospect. They experience a frictionless buying process, get immediate value, share the tool with their internal team, and invite external collaborators. The adoption loop feeds itself, and AI simply accelerates how fast that loop spins.

Value Versus Impact

The final piece of the AI revenue machine is how you communicate, and most B2B companies get this completely wrong. They all sound exactly the same - saving time, reducing costs, boosting productivity. They focus entirely on the value of their product, but value is simply what your product can do. It’s a feature. If you sell a car, the value is that it drives fast or uses electric energy. The impact is entirely different: the impact is that the car allows you to visit your mother who lives a hundred miles away.

Jacco summarized this dynamic:

“The impact is an outcome, the value is a promise. We often say value is the promise of impact. Impact is the realization of that promise.”

When you sell value, you’re just making a promise. Buyers don’t care about your promise - they care about their own reality. This means you have to completely shift your messaging. Look at what your early customers actually achieve with your product. Listen to the exact words they use to describe their experience.

And here’s where most marketing teams ruin everything: they take raw, powerful customer feedback and try to make it sound professional. They sanitize it. They polish it until it sounds “clean and sexy.” When you do that, you lose the touch - you lose the exact verbiage that actually resonates with your buyer. Don’t polish your copy until it sounds like every other software company. Use the exact words your early customers use in your marketing campaigns. Keep it in their natural language.

When you are trying to differentiate your brand in a noisy market, you will have to make hard choices about which pain point to attack and which messaging to scale. This is where the human element becomes your ultimate advantage in an AI world. You can’t just wait for a dashboard to tell you what to do - you have to rely on your intuition to find the starting point.

Jacco van der Kooij left me with one final piece of advice that I think captures the whole conversation beautifully:

“Your gut tells you where to look for the data, then the data will narrow it down and make sure that you get it right.”

Trust your gut to find the initial direction. Then use your data architecture to prove it’s actually right. This is how you win the next era of go-to-market.

Live stream: Clay for Growth & Marketing

On March 19th, I was joined by Davide Grieco, Head of Growth at Clay. He walked us through how growth and marketing teams are using Clay to move faster, act on signals in real time, and finally get credit for the ROI they generate.

We demoed real playbooks - from auto-refreshing ad audiences and signal-based campaigns to personalized landing pages at scale - the exact workflows teams at OpenAI, Vanta, and Verkada are running today.

If you’re tired of slow campaigns and budget justification loops, watch the replay of the live stream.

Watch here

Launching a product is hard.

You do not have to guess your next steps.

I built the ultimate 100-Step Go-To-Market Checklist to help you launch with total confidence and zero chaos. It includes proven website templates, sales decks, landing pages, outbound sequences, LinkedIn post frameworks, email sequences, and 20+ workshops you can immediately run with your team.

It covers every single phase of your go-to-market motion.

→ Customer research and positioning

→ Pricing and packaging strategies

→ Distribution and growth loops

This is the exact step-by-step framework I use to help my clients scale.

Stop guessing and start executing.

Get the GTM Checklist

This newsletter is sponsored by Swan AI.

AI without context is just autocomplete with confidence. It sounds right, moves fast, and produces nothing you’d actually send.

Swan built 6 FREE skills - one for every stage of the GTM motion - that give your AI agent the methodology and context it’s missing. Works in Claude, Swan AI, or any AI agent.

Get the GTM AI Agent Playbook

Dear GTM Strategist!

It is time to revisit one of the most controversial themes in our space: GTM Engineering.

Over the last year, the discussion has been heated. Some claim the title exists because GTM tech is broken. Others argue that GTM Engineers are simply tool jockeys with more love for automation than for real business fundamentals. There are even voices suggesting the entire role is vendor hype designed to sell more software.

Despite all that, the demand keeps growing.

GTM Engineers are being hired across companies and continents, and many of them are earning serious money while building the infrastructure that powers modern revenue teams.

Even if the concept irritates you a little, it is difficult to argue with the data.

So what does the data actually say?

Together with Garrett Wolfe and Alex Lindahl, we ran the largest GTM Engineering Report to date to understand what this role really looks like in practice.

Thanks to our wonderful partners who supported the report: Rho, Wiza, Juicebox, Watt Data, and Exportly.

• 228 respondents

• 30+ countries represented

• The first benchmark study of compensation, tooling, and org design for GTM Engineers.

Some highlights from the data:

• US median compensation sits around $135K

• 72% report direct revenue impact

• 84% of GTM engineers use Clay

• Nearly 68% hold no meaningful equity

• Coding skills create a $40K compensation premium

Thank you to everyone in the community who contributed data, experience, and insights. This report would not exist without you 🙏.

I want to walk through the findings and answer the questions that keep coming up in every conversation about GTM Engineering:

• How much can a GTM Engineer actually earn?

• Is this role just an agency hype, a Clay trend, or is it becoming core infrastructure for in-house teams as well?

• Do you really need to know how to code to get into GTM engineering?

• And which tools are actually powering this new wave of revenue builders? Because I do love tools as much as you do.

Now, let’s dive into the weeds of GTM Engineering.

How Much Do GTM Engineers Actually Earn?

The first thing everyone asks about GTM Engineering is simple.

Is this a real role with real compensation, or just another fancy title for someone building outbound workflows?

The data makes one thing very clear. This is a very real role and the compensation is already significant.

Across the 228 respondents in our survey, the median base salary for in-house GTM Engineers in the United States sits at around $135K.

But the distribution is extremely wide.

At the lower end, salaries fall between $60K and $90K. These are typically junior operators focused on tooling execution, list building, and basic automation workflows.

At the top end of the market, compensation climbs well beyond $200K, especially for operators who combine technical depth with ownership of pipeline generation and revenue systems.

This spread reveals something important.

The market has not figured out how to price this role yet.

Two people with the same title can be doing completely different work. One might be managing enrichment workflows and outbound tooling. Another might be designing signal engines, building automation layers, and architecting the entire revenue data stack.

Those roles are not even remotely comparable, yet the market keeps calling both of them “GTM Engineers.”

The geographic compensation gap is equally striking.

While US-based operators report a median base salary of $135K, their non-US peers sit closer to $75K. That is a difference of roughly $60K, or an 80% premium for US talent.

Even in an increasingly remote world, location still plays a major role in compensation.

Company maturity also matters.

Mid-size companies with 201 to 1,000 employees currently pay the highest median salaries for GTM Engineers. Funding stage also influences compensation, with Series B and Series D companies leading the market at roughly $145K median base salary.

Early-stage startups tend to offer lower base salaries, with pre-seed and seed companies averaging closer to $85K, often expecting equity to compensate for the lower cash component.

Except that is not what the data actually shows.

Equity distribution across GTM Engineers is surprisingly low, even in early-stage environments. Nearly 68% of respondents report holding little or no meaningful equity, despite directly influencing pipeline generation and revenue outcomes.

In other words, companies are asking these operators to build their revenue engines without always giving them a meaningful share of the upside.

This mismatch becomes even more obvious when you look at the technical side of the role.

Because there is one factor that consistently moves operators into the highest compensation bands.

Coding. And it makes me wonder - is the next GTM system a structure of folders and GTM as a code?

Do GTM Engineers Need to Know How to Code?

The short answer is no. You can absolutely enter the field without coding skills. But you are likely to earn much more if you do know how to code.

Our data shows that technical depth has a very real impact on compensation and scope.

When we analyzed the survey responses, a clear pattern emerged. GTM Engineers fall into three broad categories based on their technical proficiency.

1. Low-code operators.

2. Mid-level technical builders.

3. High-code engineers.

Low-code operators typically rely on no-code platforms and workflow builders. They connect tools, manage enrichment pipelines, and run outbound programs using existing platforms. Their work focuses on configuration rather than engineering. Typical tools in this category include Clay, Zapier, Make, Airtable, and no-code CRM automation inside HubSpot or Salesforce.

These operators report a median salary of around $90K.

The next group sits somewhere in the middle. These professionals understand scripting, basic APIs, and data manipulation. They might write small pieces of code, automate tasks, or build lightweight integrations when needed. They often combine platforms like Clay or HubSpot with scripts written in Python, JavaScript, or SQL to extend what existing tools can do.

Their median compensation rises to roughly $105K.

Then there is the top tier.

High-code GTM Engineers regularly work with Python, JavaScript, SQL, and modern AI coding tools. They build custom workflows, write scripts, create data pipelines, and sometimes bypass traditional vendors entirely by building their own internal tooling. Tools in this group often include Python environments, data warehouses like Snowflake or BigQuery, and AI-assisted development tools such as Cursor or Claude Code.

This group reports a median salary closer to $135K.

That is roughly a $40K to $45K premium compared to low-code operators.

In practice, the role is splitting into two different types of operators.

Some GTM Engineers focus primarily on operating the modern sales stack. They configure tools, manage enrichment workflows, and run outbound systems using existing platforms.

Others spend most of their time building the infrastructure behind those systems. They write scripts, build custom automations, and design the data pipelines that power the go-to-market engine.

Both roles are valuable, but they require very different skill sets and experience levels.

Interestingly, the rise of AI-assisted coding is already changing the skill requirements. Tools like Cursor and Claude Code allow GTM Engineers to generate scripts, build automations, and experiment with custom workflows without needing the background of a traditional software engineer.

This does not eliminate the value of technical knowledge. If anything, it increases the leverage of operators who understand how systems work.

But the data suggests that the closer you move toward engineering, the more valuable and better compensated your role becomes.

GTM Engineering in Agencies vs In-House Teams

One of the most common questions around GTM Engineering is where this role actually belongs. Is it something companies should build internally as your core capacity - if so, where should GTME sit in an org, or is it better suited for agencies and external specialists to test the grounds?

The data suggests that both models exist and are growing quickly, but they operate very differently. A large part of the early momentum around GTM Engineering came from the agency ecosystem.

Over the past two years, a wave of specialized agencies and fractional operators started offering GTM engineering services. The barrier to entry was relatively low, demand from founders was extremely high, and the tooling ecosystem made it possible for a single operator to build fairly complex outbound infrastructure.

Our data reflects this rapid growth and the chaos that comes with it.

Monthly agency retainers range from as little as $1K to as high as $33K per month. The median minimum fee sits around $5K, while the median maximum retainer lands closer to $8K.

This is an enormous spread.

The reason is simple. Buyers and sellers are often talking about completely different services while using the exact same label.

Some agencies provide done-for-you outbound. They build lists, run enrichment, and launch campaigns.

Others operate as systems architects, designing signal engines, building data pipelines, and integrating the entire revenue stack.

These are fundamentally different services.

Yet they are both marketed as GTM Engineering.

At the same time, the role is increasingly appearing inside companies themselves.

A growing share of respondents work in-house inside revenue teams, not as consultants or agencies. These operators sit directly within marketing, RevOps, or growth teams and own the internal automation layer that connects the entire go-to-market motion.

Instead of running campaigns for multiple clients, they focus on a single company and build long-term infrastructure.

• Signal monitoring.

• Pipeline automation.

• Data enrichment systems.

• CRM architecture.

This is where the role becomes far more strategic.

In-house GTM Engineers are not just launching outbound sequences. They are designing the systems that determine how pipeline is generated, routed, enriched, and converted across the entire funnel.

Agencies often help companies get started. In-house teams are where these systems are maintained and scaled over time.

So while agencies helped popularize the category, the role itself is quickly becoming core infrastructure inside modern revenue teams.

The most mature organizations are not outsourcing this capability.

They are building it internally.

The GTM Engineering Tool Stack

If you want to understand a new role, look at the tools the operators actually use.

The GTM Engineering stack reveals a lot about how modern revenue teams are evolving.

CRM platforms remain the foundation. Salesforce and HubSpot still dominate the ecosystem, with roughly 88% adoption among respondents. Almost every GTM Engineer is building automation layers on top of one of these systems.

But the real story sits above the CRM.

Clay has quickly become the default infrastructure for this function. Across the survey, 84% of GTM Engineers report using Clay, making it the most widely adopted tool in the entire stack.

Among agencies the adoption climbs even higher, reaching 96%.

This level of dominance is rare for such a young platform.

Clay essentially acts as the connective layer between enrichment providers, data signals, outbound workflows, and automation logic. For many operators it has replaced large portions of the traditional outbound stack.

At the same time, Clay is also the most polarizing tool in the ecosystem.

Operators love the flexibility. They can combine dozens of data sources, trigger workflows based on signals, and build highly customized outbound engines.

They also complain about the complexity.

Workflows often rely on multiple APIs, fragile integrations, and complex logic chains. Running a single campaign can require stitching together several tools and enrichment providers. When one piece breaks, the entire workflow stops.

This creates a strange paradox.

Clay is the most powerful platform in the stack and also one of the most frustrating to maintain.

Meanwhile, AI-native development tools are spreading quickly across the GTM engineering community.

Tools like Cursor and Claude Code now show adoption rates approaching 70% among respondents. These tools allow operators to generate scripts, build small internal tools, scrape data, and automate workflows without relying entirely on third-party vendors.

This shift toward AI-assisted development is changing how GTM Engineers operate.

Instead of waiting for vendors to build new features, many operators simply build their own solutions.

When we asked respondents what they value most in their tools, the answers were very consistent.

• 44% pointed directly to AI capabilities.

• Another 24% highlighted automation power and speed.

At the same time, the frustrations are equally clear.

• 26% complain about poor integrations and closed ecosystems.

• 18% struggle with clunky interfaces.

• 11% cite expensive platforms that deliver limited flexibility.

Despite the massive popularity of Clay and the rapid rise of AI tooling, a clear gap still exists in the market.

About 12% of respondents say they wish a true all-in-one outbound platform existed.

The demand for a unified system is obvious.

So far, no vendor has fully solved it.

For now, GTM Engineers continue to build their own stacks.

The Maturity Gap and the Hype Train

GTM Engineering might be one of the fastest-growing roles in modern revenue teams.

It is also one of the youngest.

When we analyzed the age distribution of survey respondents, the results were surprising.

Nearly 30% of GTM Engineers are between 18 and 22 years old. Another 25% fall between 23 and 26.

That means more than half of the talent pool entered the workforce only a few years ago.

This explains a lot of the confusion surrounding the role.

Many operators entering the space are extremely strong technically. They know how to use modern sales tech. They can build workflows, connect APIs, and automate data enrichment pipelines.

But deep experience with complex sales processes is still relatively rare.

That experience gap matters.

GTM Engineering sits directly at the intersection of marketing, sales, RevOps, and data infrastructure. The operators building these systems are shaping how pipeline flows through the entire organization.

When someone with limited business experience is suddenly responsible for designing revenue systems, mistakes are inevitable.

This does not mean the role is flawed.

It simply means the category is still very early.

The tooling ecosystem exploded quickly. Clay, enrichment APIs, signal platforms, and AI coding tools dramatically lowered the barrier to building automation systems. Suddenly, a single operator could do work that previously required an entire RevOps team.

Naturally, a wave of ambitious builders rushed into the space.

Some are incredibly talented. Others are still learning the fundamentals of how revenue organizations actually operate.

This creates a chaotic market.

Companies are hiring junior tool operators and expecting them to perform like senior revenue architects. Operators are experimenting aggressively without always understanding the long-term implications of the systems they are building.

The result is a wide maturity gap across the industry.

Some GTM Engineers are quietly becoming some of the most valuable operators inside modern revenue teams.

Others are still figuring out what the role actually requires.

This is exactly what early categories look like.

Messy. Fast-moving. And full of experimentation.

Where GTM Engineering Goes From Here

For all the discussion about tools, AI, and automation, the biggest constraint facing GTM Engineers is surprisingly simple.

Time.

In our survey, 25% of respondents said bandwidth is their number one bottleneck. Most GTM Engineers operate as a team of one, responsible for designing automation systems, maintaining the data layer, supporting sales workflows, and experimenting with new go-to-market infrastructure.

At the same time, many companies are still figuring out what this role actually is.

Only 45% of respondents say their organization clearly understands what a GTM Engineer does. Another large group reports only partial understanding across leadership and revenue teams.

This combination creates an interesting moment.

The tools are evolving quickly. The talent pool is growing. But the organizational structure around the role is still catching up.

The companies that figure this out early will have a real advantage.

Because modern go-to-market is becoming increasingly technical. And the operators who can connect systems, automate workflows, and build revenue infrastructure are quickly becoming some of the most valuable people inside revenue teams.

So there you have it. A snapshot of where GTM Engineering stands today, what the market pays for it, and which skills will make you indispensable in the years ahead. If you want more information, download the full The State of GTME Report here. What I shared in this article are my key takeaways, the report itself is 80-slides long - and you can get it FREE:

Get the full report

✅ Need ready-to-use GTM assets and AI prompts? Get the 100-Step GTM Checklist with proven website templates, sales decks, landing pages, outbound sequences, LinkedIn post frameworks, email sequences, and 20+ workshops you can immediately run with your team.

📘 New to GTM? Learn fundamentals. Get my best-selling GTM Strategist book that helped 9,500+ companies to go to market with confidence - frameworks and online course included.

📈 My latest course: AI-Powered LinkedIn Growth System teaches the exact system I use to generate 7M+ impressions a year and 70% of my B2B pipeline.

🏅 Are you in charge of GTM and responsible for leading others? Grab the GTM Masterclass (6 hours of training, end-to-end GTM explained on examples, guided workshops) to get your team up and running in no time.

🤝 Want to work together? ⏩ Check out the options and let me know how we can join forces.

As a trained marketer and someone who spent 10 years working as Marketing Manager and CMO before ever having to learn RevOps/Sales for my own business, I have a hot take ♨️ on how nearly all cool technologies GTM technologies are tailored to “sales and RevOps” while marketing is massively lagging behind GTM AI transformation.

When Kyle Poyar and I launched The 2026 State of AI in GTM, I was shocked at how marketing GTM AI plays were mainly focused on productivity - how to do the exact same thing we were doing for the last 10 years faster, allegedly better, and in larger quantities. Some plays were more advanced, but so vendor-specific that I was gagged by how we lack imagination and ideas for reinventing the marketing process with the powerful AI tools we’ve had at our disposal for the last couple of years.

And I am not blaming marketing here for a sec - marketers are among the hardest working people in any organization. They run campaigns, take care of the brand, manage relationships with many stakeholders, organize events, manage advertising budgets, and handle the creative side of businesses. They were thrown AI at them and told: “Now be more productive and create better ROI.”

With increasing pressure on opex and an “adopt or die” mindset to AI, those people started to panic. FOMO on tools, feeling like their department is falling behind, and overselling on demos - best they could do was to create a couple of projects in ChatGPT/Claude to speed up content creation, bought some tools for mass generation of creatives, and at best started playing with some agents to be able to prove to their bosses that they too are AI-transformed now.

But that created problems - deeper problems that we’d expect. As a result of local AI implementations, new silos emerged within the organization. And it is even worse when you go out there and see them in person.

Mainstream B2B marketers are still sold on the idea that their job is to do “inbound marketing” - a concept HubSpot pioneered in 2006. People, that was exactly 20 years ago. This idea is dead. In 2026, we should be holistically thinking about integrated buyer experiences and ensuring that Marketing and “Sales” are elegantly aligned to drive demand toward business outcomes. And that means outbound, data enrichment, and the notion of redefining what we call Marketing Qualified Lead (MQL) with signals we have at our disposal beyond “downloaded an ebook”, “liked our LinkedIn post”, and “registered for a webinar”.

People don’t like it when I say that marketers should move into the role of business developers, sharing tools with RevOps, but that is exactly how we enable marketers to squeeze more out of the activities they are already putting so much time, money, and effort into. By working hand in hand with Marketing leaders in my clients’ orgs and running 50+ Clay tables for them, I put together a starter pack of 8 plays that show you how to use Clay across your entire marketing stack.

These aren’t edge cases or anything new to add to an already full to-do list. These are plays marketing teams are running right now, and this is how you squeeze more value out of them. People love them - I published an earlier version on LinkedIn, and the post went viral.

In this article, I’ll walk you through the concept behind each play and when to use it. For the full step-by-step implementation guide with prompts, scoring formulas, and integration details, grab the complete playbook here.

8 Marketing Clay Plays

Here’s what we’ll cover:

• Play 1: Newsletter Audience Intelligence — Turn a flat email list into segmented, personalized content

• Play 2: Competitive Ad Intelligence — Pull ad spend and messaging data across 50 competitors in one table

• Play 3: Event & Conference Intelligence — Walk in prepared, walk out with scored follow-up lists

• Play 4: Influencer & Creator Vetting — Vet 200 creators by content relevance, not follower count

• Play 5: Community & Forum Intelligence — Capture competitor complaints and buying intent from Reddit, automatically

• Play 6: Partnership & Co-Marketing Discovery — Replace “who should we partner with?” gut feelings with data

• Play 7: Programmatic Personalized Landing Pages — Generate 1:1 pages per account at scale

• Play 8: Podcast & Speaker Pitch Lists — Pitch 200 shows in the time it takes to manually research 10

Let’s dig in 🔎

Play 1: Newsletter Audience Intelligence and Segmentation

Marketing function: Email / Content Marketing

Difficulty: 2/5

You’re sending the same newsletter to everyone. Your VP of Marketing subscribers get the same content as your RevOps ICs. Your SaaS readers see the same topics as your e-commerce readers. And you have no idea who any of these people actually are — you have email addresses, maybe first names, and that’s it.

The play: Import your subscriber list into Clay, enrich every subscriber with firmographic and professional data (company, title, seniority, industry, revenue band), then use an AI column to cluster your audience into meaningful segments — not by demographics, but by content needs.

Think segments like “SaaS Leaders” (VP+ at SaaS companies, cares about strategy), “Startup Operators” (any role at sub-50-person companies, needs tactical how-tos), or “Agency/Consultants” (needs frameworks they can apply to clients).

Three outputs from one table:

• Segmented email lists synced back to your email platform — different content for different audiences

• Content intelligence — you know exactly which topics each segment cares about, so your editorial calendar is driven by data, not guesses

• Sales-ready leads — subscribers who match your ICP, already warm, exported to CRM with full context

The marketers getting this right are seeing 45%+ open rates on segmented sends. Same list. Same effort. Different results — because they know who’s reading.

Bonus: If you sell sponsorships, the enriched data is a goldmine for pricing. “38% of our subscribers are VP+ at SaaS companies doing $50M+ in revenue” justifies 2-3x higher CPMs than “we have 10,000 marketers on our list.”

Play 2: Competitive Ad Intelligence

Marketing function: Advertising / Competitive Intel

Difficulty: 2/5

Be honest — here’s how most marketing teams do competitive ad research: Check Facebook Ad Library. Screenshot a few ads. Check LinkedIn. Maybe Google Ads Transparency Center if you’re thorough. Do that for competitor #2, #3... and by #4 you’re bored and skipping. You end up with a folder of screenshots and zero pattern analysis.

The play: Build a Clay table with 50 competitors and pull structured ad data using Clay’s native integrations — Adbeat for display and video ads across 80+ networks, Semrush for paid search intelligence and traffic sources. Then use an AI column to classify messaging themes across the entire landscape.

This is where it gets interesting. Instead of knowing what one competitor’s latest ad looks like, you can answer questions like:

• What messaging themes are overused in our space? (If everyone says “Scale faster with AI,” maybe don’t.)

• What pain points are under-addressed? (Nobody’s talking about implementation pain? That’s your angle.)

• What offer types dominate? (If everyone offers demos, maybe a free trial stands out.)

• Who’s spending the most and where? (Actual budget benchmarks, not guesses.)

Output: A competitive positioning brief with real spend data, messaging pattern analysis, and 3 specific recommendations for how to differentiate. Updated whenever you want — not once a quarter.

Good news - Clay just launched an Ads feature so you can build ad audiences directly in Clay, enrich them, and with that achieve a much better match rate (up to 90%, compared with standard 30% if you work only with firmographics) - and drastically lower your cost-per-lead. Then you can push them directly to LinkedIn Ads or Meta Ads using the Clay integration.

Watch my Clay Ads demo:

Play 3: Event and Conference Intelligence

Marketing function: Events / Field Marketing

Difficulty: 3/5

Events are the biggest line item in most marketing budgets. And the follow-up is always the weakest link.

Pre-event, you LinkedIn-stalk 20 people the night before and hope for serendipity. Post-event, badge scans sit in a spreadsheet and someone sends a generic “Great meeting you!” email two weeks later. You spent $30,000 on a booth and can’t prove ROI.

The play has two halves:

Before the event: Import the attendee/speaker list into Clay, enrich with company data, then — and this is the key step most people skip — define your trigger signals and scoring priorities. Not generic firmographics. Your signals. Are they hiring a RevOps person (signal: building the team that buys your tool)? Using a competitor’s product? Recently raised Series B? New CMO in the last 90 days?

Use a Claygent column to research each attendee against your custom triggers. Then generate personalized talking points that reference their actual situation.

Instead of walking up with generic “How’s your team scaling?”, you get “I saw you’re hiring a RevOps lead and just moved off [Competitor] — how’s the transition going?” That starts a real conversation.

After the event: Import engagement data (badge scans, conversations, demo requests) and score attendees by interaction depth. A badge scan gets a generic nurture. A demo request gets AE handoff the same day. Everyone in between gets follow-up matched to their engagement level, with messaging that references their specific situation.

Output: Walk in prepared with a ranked “who to meet” list. Walk out with prioritized follow-up that hits within 24 hours.

Play 4: Influencer and Creator Vetting at Scale

Marketing function: Influencer Marketing

Difficulty: 3/5

B2B influencer marketing is exploding, and vetting is 90% of the work. Find a creator. Go to their LinkedIn. Scroll through posts. Check engagement. Are the comments real? Do they actually post about relevant topics? Now do that 50 times.

Most teams vet 5-10 creators manually and pick based on gut feel and follower count. Follower count is a vanity metric. A creator with 5,000 engaged ICP followers beats one with 100,000 random ones every time.

The play: Feed 200 potential influencers into Clay. Use enrichment providers (Apollo, Clearbit, Modash) for structured profile data — title, employer, follower counts, engagement rates. Then use Claygent for thought leadership research — searching the open web for their articles, podcast appearances, conference talks, and blog posts.

Score by content relevance (do they write about your topics?), professional credibility (what’s their background?), content volume (are they actually active?), and platform reach (weighted last, not first). Add a red flag detection layer — are they employed at a competitor? Have they gone inactive?

For everyone scoring above threshold, generate personalized pitches that reference their specific content. Not “we’d love to leverage your audience” but “Your article on signal-based selling aligns with what we’re building — here’s a specific collab idea.”

Output: A data-driven influencer selection in hours, not weeks. Decisions based on what creators have actually published, not their follower count.

Play 5: Community and Forum Intelligence

Marketing function: Community / Competitive Intel

Difficulty: 1/5

Reddit threads, industry forums, public community discussions — people are literally typing out their problems, frustrations with competitors, buying criteria, and feature requests. In public. For free. And nobody monitors them systematically.

The play: Use Clay Signals (a free, native feature) to automatically monitor Reddit for mentions that match your criteria — competitor names, pain-point phrases (”looking for a tool that…”, “frustrated with…”, “alternative to…”), and buying intent signals. When Signals flags a relevant discussion, use Claygent to visit the specific thread and extract deeper intelligence.

Each mention gets classified by signal type:

• Competitor complaint → Feeds your battle cards and competitive positioning

• Pain signal → Becomes your next blog topic or ad angle

• Feature request → Goes to the product team

• Buying intent → Alerts your content team or community manager to engage (with a helpful comment, not a sales pitch)

• Market insight → Monthly intelligence digest for the team

This is a market intelligence play, not a lead generation play. You’re capturing what the market is saying to inform your marketing — not identifying individuals behind anonymous posts.

Output: A continuously updated competitive and market intelligence feed that runs on autopilot. Monthly digests with top competitor complaints, trending pain points, and content topics the market is actively asking about.

Play 6: Partnership and Co-Marketing Discovery

Marketing function: Partnerships / BD

Difficulty: 3/5

“Who should we partner with?” is a question every marketer asks. Nobody has a system to answer it. You scroll LinkedIn, think of companies your customers also use, ask around, and hope someone has a warm intro. You end up partnering with whoever said yes, not whoever would be the best fit.

The real question isn’t “who should we partner with?” — it’s “who has our audience but doesn’t compete with us?” That’s a data question.

The play: Start with a real data source — account overlap from Crossbeam (which companies share customers with you?), AI lookalikes from Ocean.io (which companies serve the same buyer in adjacent categories?), or tech stack matches from BuiltWith (whose customers use similar tools?).

Import candidates into Clay and enrich with compatibility signals: Are they actively marketing (blog, newsletter, webinars)? What’s their audience size and social presence? Are they a similar company stage? Use Claygent to visit their website and assess their marketing activity level.

Score by partnership compatibility — audience overlap, complementary offering, size match, marketing activity level — and generate personalized co-marketing pitches for the top-scoring candidates. Not “let’s partner” but a specific idea: “Let’s co-create a benchmark report on [shared topic] — your analytics audience and our automation audience would both benefit.”

Output: A scored partner list built from real data, not guesswork, with pitches ready to send.

Play 7: Programmatic Personalized Landing Pages

Marketing function: Content / ABM

Difficulty: 4/5

“Personalized landing pages” has become meaningless. Most teams swap the company name and logo on a template and call it personalization. “Welcome, Acme Corp!” and then the same page for every visitor. Nobody falls for it.

Real personalization means different messaging for different accounts — industry-specific value props, relevant case studies, company-specific challenges, tailored CTAs. But creating truly personalized pages is brutal at scale. 50 target accounts × 1 custom page each = 50 pages nobody has time to write.

The play: Start with your target account list in Clay. Enrich with company data and use Claygent to research each account’s likely business challenge (from job postings, recent news, public strategy signals). Then generate page copy per account — headline, value proposition, case study match, and CTA — all tailored to that account’s situation.

Push everything to Webflow’s CMS using Clay’s native integration. One page template in Webflow with dynamic fields. Clay populates the fields differently per account. Each gets a unique URL: yoursite.com/for/[company-name].

Use these pages across channels — ABM campaigns, event invitations, partnership proposals, outbound sequences. Track per-page engagement and alert your team when accounts show high interest.

Output: 50+ landing pages where the content actually IS different. Not “Hello {company_name}” mail merge. Real relevance — a fintech company sees a compliance-focused value prop and a finserv case study, while a SaaS company sees a scaling-focused message with a matching case study.

Play 8: Podcast and Speaker Pitch Lists

Marketing function: PR / Business Development

Difficulty: 2/5

Podcast guesting is one of the most underrated marketing channels. One good appearance puts you in front of a captive audience for 30-60 minutes. No other channel gives you that kind of attention.

But podcast outreach is 95% research and 5% pitching. Find a podcast. Check if it’s still active. Find the host. Find their email. Listen to an episode. Write a pitch. At that pace, you pitch 5 podcasts a week if you’re dedicated.

The play: Start with a real podcast database — Listen Notes has 3.7M+ indexed shows searchable by topic, with structured data on host, episode count, and popularity. Export to CSV and import into Clay. Then use Claygent to research each podcast’s website — host names, guest booking pages, recent episode topics, typical guest caliber. Use enrichment providers (Apollo, Hunter, Findymail) to find host emails.

Score by topic relevance, guest caliber match (do they host people at your level?), recency (active vs. dormant), audience signal, and contact availability. For every podcast above threshold, generate a personalized pitch that references their specific recent episodes and proposes 2-3 concrete topic ideas.

Output: 200 podcasts researched, scored, and pitched in the time it takes to manually research 10. Each pitch references the host’s actual content, not a generic “I’d love to be on your show.”

Where to Start

Unlike ABM plays that build on each other, these marketing plays are independent. You don’t need Play 1 to run Play 4. Pick the ones that match your biggest pain:

• Spending too much on events with weak follow-up? Start with Play 3.

• Newsletter engagement is flat? Start with Play 1.

• No idea what competitors are saying in their ads? Start with Play 2.

• Need influencers but can’t find the right ones? Start with Play 4.

My recommended order: Start with Newsletter Intelligence + Competitive Ads (week 1 — fastest to build, you already have the data). Then move to Event Intelligence + Influencer Vetting (week 2-3 — high-ROI plays affecting big budget lines). Add Community Intel + Partnerships (week 4-5 — signal plays that compound over time). Then tackle Landing Pages + Podcasts when you’re ready for more complex builds.

There are also natural synergies between the plays — community intelligence (Play 5) reveals pain points that inform your competitive positioning (Play 2), which feeds your personalized landing pages (Play 7). Event attendees (Play 3) enrich your newsletter intelligence (Play 1). Influencer vetting (Play 4) surfaces creators who also make great podcast guests (Play 8).

Start with 2-3 plays. Build the muscle. Add more as you see results.

And do it with patience and love. Some marketers might be a bit overwhelmed at first - set them up for success - double down on what is already working well for them and extract leads from that. Help them help you should be the guiding principle here. They are not “less likely to understand our tools” - they are just 2 years behind in adopting them, and they can catch up really fast. I’ve seen it work. Hope you’ll do it too

For the full implementation guide with step-by-step instructions, Clay prompts, scoring formulas, and integration details for all 8 plays, grab the complete playbook here.

If you like what I do, you can find more of such assets, playbooks, workshops, and awesome examples on my 100-Step GTM checklist.

This post is supported by Schematic — monetization infrastructure for product & engineering teams.

Schematic decouples billing from code, giving product teams direct control over pricing, packaging, limits, add-ons, credits, trials, migrations, and upgrades — without burdening engineering.

The result: faster pricing experiments, better conversion, more upgrades, and less revenue leakage. Works natively with Stripe Billing. Used by high-growth teams like Plotly, Pagos, TightKnit, and Zep.

Try Schematic free

Dear GTM Strategist!

Earlier this year, Kyle Poyar and I published our 2026 State of AI for GTM report. The results were brutal. 53% of GTM leaders reported little to no impact from AI. Only 24% are seeing real returns. Nearly half don’t have a single AI agent in production.

Why is the gap so big? It isn’t the models. We all have access to the same LLMs. It isn’t prompt engineering. We all have the same cheat sheets.

The difference is structural. The 53% are treating AI like a chatbot. One-off conversations. Zero memory. Starting from scratch every time. The 24% are building Context Systems.

This shift has a name: Context Engineering.

It is the practice of designing the entire information environment around your AI, not just the prompts you type into it. It is the difference between asking AI clever questions and building a system that actually understands your business.

We are seeing a massive rise in GTM Engineering teams adopting this approach. And right now, the primary environment for building these systems is Claude Code.

In this post, I’m breaking down:

1. What context engineering is and why it changes the AI equation for growth teams

2. How Claude Code makes it practical, even if you’ve never opened a terminal

3. The four building blocks that turn AI from a chat window into a compounding knowledge system

4. Where to start (simpler than you’d think)

The Problem: AI Without Context Is Expensive Autocomplete

Here’s what actually happens in most companies right now.

A marketing manager opens ChatGPT. Types “write me a cold email for our product.” Gets back something that sounds polished but could have been written for any B2B SaaS company on earth. No understanding of the ICP. No awareness of the positioning. No knowledge of what worked before and what flopped.

So they try to fix it. “Make it more specific to fintech CFOs.” Better, but still generic. “Add a reference to regulatory compliance challenges.” Getting warmer, but the tone is off. “Make it sound less like AI wrote it.” Now it’s bland. “Actually, go back to version two but keep the compliance angle.” Five, eight, twelve rounds of back-and-forth, each prompt trying to inject a piece of context that the AI should have known from the start.

After 30 minutes of this prompt ping-pong, you end up with something that’s... acceptable. Not great. Acceptable. The next time you need an email? The exact same process. From zero. All that refinement you just did? Gone. The AI learned nothing from the exchange.

Most growth and marketing teams live in this loop, and it’s the reason behind those survey numbers. Every interaction starts from scratch. No memory, no accumulated knowledge, no understanding of your business. You’re re-onboarding a team member every single conversation. Except this team member has amnesia and forgets everything the moment you close the tab.

Imagine you hired a senior marketer, a really good one. Instead of briefing them once and building on that knowledge over time, you wiped their memory clean every morning and started over. You’d fire that person in a week.

Most teams are doing exactly that with AI. Then wondering why 53% see no impact.

Enter Context Engineering

In mid-2025, Shopify CEO Tobi Lütke posted something that caught the attention of the entire AI world:

“I really like the term ‘context engineering’ over prompt engineering. It describes the core skill better: the art of providing all the context for the task to be plausibly solvable by the LLM.”

Andrej Karpathy, one of the most respected AI researchers alive (former Tesla AI lead and OpenAI founding member), immediately agreed:

“Context engineering is the delicate art and science of filling the context window with just the right information for the next step.”

These aren’t fringe voices. The CEO of a $100B+ company and one of the most cited AI researchers in the world are saying the same thing: the way we work with AI is about to look very different.

Here’s how the shift breaks down:

Prompt engineering was about crafting the perfect question. Getting your words just right. Finding the magic sentence that made the AI produce good output. Focused on the input, one message at a time.

Context engineering is about designing the entire information environment around the AI. Not just what you ask, but what the AI already knows when you ask it. Your strategy, your ICP, your brand voice, your past learnings, your tool stack. Structured, persistent, and available at the moment it’s needed.

The simplest way to see the difference:

Prompt engineering is writing a clever brief for every single task. Context engineering is building the team that doesn’t need a brief, because the institutional knowledge, the playbooks, the quality standards, and the tool access are already baked into how the operation runs.

One scales with effort. The other scales with knowledge.

Why Growth and Marketing Teams Should Pay Attention

If you work in growth, marketing, or GTM operations, context engineering isn’t an abstract technical concept. It’s the answer to a specific, expensive problem you’re probably already experiencing.

The problem: Your team uses AI for content creation, email copy, competitive research, lead scoring, or campaign analysis. The output is fine. Usable. But generic. It “sounds like AI.” You end up rewriting half of it to match your voice and add the specifics only you would know.

The root cause: The AI doesn’t know your business. It doesn’t know your ICP is mid-market B2B SaaS companies with 50-200 employees who just raised Series A. It doesn’t know your brand voice is direct and practitioner-first, not corporate and safe. It doesn’t know that your outbound campaigns perform better when you lead with the industry-specific pain point instead of the feature list.

All of that context exists somewhere: in your head, in scattered docs, in your CRM, in the lessons your team has learned through trial and error. But the AI can’t access any of it. So it defaults to generic. Every single time.

Context engineering fixes this by building persistent structures that give AI access to the right knowledge at the right moment. Not dumping everything into a prompt. Not copying and pasting your brand guide at the start of every session. A system that loads the right context automatically and accumulates knowledge over time.

The 24% of teams seeing real AI impact and the 53% still getting nothing are using the same underlying models. The difference is the infrastructure around them.

Claude Code: Where Context Engineering Gets Practical

Let’s get tangible.

Claude Code is Anthropic’s command-line tool for working with Claude. If you’ve only used AI through chat interfaces (ChatGPT, the Claude website, Gemini), Claude Code works differently. That difference is the whole point.

It runs in your terminal (the text-based interface that developers use). If you’ve never opened a terminal, that might sound intimidating. What matters for growth and marketing teams: Claude Code is the first AI environment actually designed for context engineering.

In a normal chat interface, you type, the AI responds, and eventually the conversation gets long enough that the AI starts “forgetting” earlier details. No persistent knowledge. No structure. No way to say “always know these things about my business before we start.”

Claude Code changes that with four building blocks that turn disposable conversations into a persistent knowledge system.

I’ll walk through each one.

The Four Building Blocks

1. CLAUDE.md — Your Strategic Brain

Your strategic onboarding document for the AI, read automatically at the start of every session.

CLAUDE.md is a text file where you write down everything the AI should know about your business. Your positioning. Your ICP definition. Your brand voice guidelines. Your pricing model. The frameworks you use. Navigation rules so it knows where to find what. How your tools connect and when to use which one.

If you’ve ever onboarded a new marketing hire, you know the drill. The first two weeks are pure context transfer: brand guides, competitive landscape, audience research, past campaign performance, what worked and what flopped. CLAUDE.md is that entire onboarding compressed into a document the AI absorbs in seconds.

What makes it powerful is persistence. Unlike a chat conversation that evaporates after you close the tab, Claude Code reads your CLAUDE.md at the start of every session. The AI never needs to be re-briefed. It shows up already understanding your business, your standards, and your preferences.

It also evolves. When you learn something new (a sharper ICP insight, a campaign angle that outperformed, a positioning shift), you update the file. Over time, the CLAUDE.md becomes a living strategic document: your institutional knowledge, accessible to every AI interaction going forward.

As your system matures, the CLAUDE.md naturally evolves from holding everything directly into a navigation layer, pointing the AI to separate, dedicated files for your ICP research, your voice guide, your competitive analysis. You start with one well-organized document and gradually build it into a structured knowledge base. The CLAUDE.md stays at the center, but the depth behind it grows.

The companies getting real ROI from AI aren’t writing better prompts. They’re building better knowledge files.