How AI products Go-to-Market
How AI products Go-to-Market
Case studies: CommandBar (UX), Rewriter Go-to-Market strategy
This Substack post is kindly supported by CommandBar, all-in-one AI-Powered User Assistance (onboarding, activation, chats, support assets, etc.) used by 20 million users and best-in-class companies like Clearbit, AngelList, ConvertKit, Gusto, 6sense and many others. See what they do here: https://www.commandbar.com
Dear GTM Strategist!
Let me reveal a secret …
I have been working on 24 AI-first projects in Q2 and I am building a new AI-based product myself (stealth mode for now, but I promise you’re going to love it).
Long story short - GTM for AI-first products is different.
In my experience, it is easier because we can ride the momentum, and almost every team that I have been working with got 1000s of registrations, 100 weekly active users (in month 3) and 50 paying clients with 0 budget (posts to groups, social media and outbound). The blueprint for how to promote these products is simple. Pricing, retention, and positioning of AI products are not.
I am working on advancing my frameworks to serve AI companies even better and I started to systemize my learnings that will primarily be presented as a case study during my keynote “What is the Go-to-Market Strategy for AI Products?” at Product Drive Summit. You are welcome to register. It is free, and the speaker lineup is amazing, as always ✌️
By now, I firmly believe that by 2025, the adoption of AI will be not just a trend but a necessity for every company. It's high time to start planning your AI strategy.
But … hooking up a random chatbot to your website and using ChatGPT to write your LinkedIn posts is so 2023.
As AI is maturing so should we.
The days of AI romanticism will soon be over.
As with anything in tech, AI solutions can be used:
To enhance productivity and increase the efficiency of existing processes by buying or developing AI solution - this is the easier place to start for most companies.
To add new features, capabilities, and business models to advance the business and develop your own solutions where it makes sense.
In this Substack, I will invite you to:
Explain AI to Martha from LinkedIn: Understand the fundamentals of AI by grasping the architecture of AI solutions, my thoughts on how the landscape is evolving.
Use case 1: Use AI solution to advance existing processes: We all know you can make a lot of money if you fix activation in your product. CommandBar’s AI-powered user assistance has some cool features that will delight users and add value instead of annoying them with pestering pop-ups.
Use case 2: Launch your own AI-product & open it to others: Analyze a case of how Rewriter (GTM PLan how to get 1000 users with 0 budget)- a LinkedIn post generator trained on 60,000 viral LinkedIn posts - attracted alpha, beta, and public beta users —free as always. I have been with the maker since alpha version and we are just getting ready to launch open beta.
3 mistakes to avoid when launching an AI Product: I end this Substack with main lessons learned from my recent AI launches. Coz why not? 🤠
We will make AI friendly for non-technical people.
You will grasp it in a sec, I promise.
For all the tech folks, let’s make you some money with your AI superpowers!
Buckle up and let’s do it!
A quick intro to the AI landscape and core issues
Let’s first align our understanding of AI. In 2024 the investments in AI solutions will continue to grow. There is a running joke among my VC friends to run “command + F” search on pitch decks and count how many times a startup mentions the word “AI” in their pitch.
AI is not a new technology. It has been around since the 1950s, some would argue even earlier (learn more about the history of AI if you are a true history geek such as I am 🤓), but ChatGPT is the first widely adopted consumer usage where users are directly exposed to AI and as such impossible to ignore. ChatGPT by OpenAI is the fastest service got 100 million users which they managed in 2 months in 2022. As of this writing, ChatGPT is estimated to have 200 million active users and over 600 million monthly website visitors.
I believe this is one of the best product-led, value-first user-inclusive examples I will ever see in my career. What made ChatGPT so successful? I’d argue that low entry barriers and time to value - the tool is free to use, produces amazing value added in seconds, and solves real pain points. And we are just getting started:
The technology behind it is called generative AI. In a nutshell, it takes a prompt (a snippet of text or other media) as input and produces a stream of output in response to and based on the given prompt. While the most common combination is text for prompt and output, that need not be the case. All of the state-of-the-art generative AI systems are multimodal, meaning they can take input and produce output in multiple formats (e.g. prompt is a picture and some textual instructions, output is synthesized speech).
What makes this iteration of AI super accessible and enticing is that while most people cannot code, but they can write prompts 🤠- especially when they can start experimenting free of charge.
While it is easy to get hyped about AI, let’s establish a basic understanding of the different layers of AI solutions. Most AI tools that you know are AI applications (ChatGPT, Jasper…). But in order for these apps to work, we need other layers too. ChatGPT is built on large language models developed by OpenAI, such as GPT-4. And in order to develop and train those, you need computing/processing power. A lot of it.
Now things get interesting. Most competition is on the top application layer, but the most profits and investments are on the bottom layer layers (infrastructure, inference, compute), which are dominated by big players such as OpenAI, NVIDIA, Google, Meta, Amazon, Anthropic - you know - the usual suspects. That creates a critical dependency in the system - if they change the terms/availability/price of their services for applications, the market changes.
We have seen this before with prices and limitations of media buying (paid ads). As Meta changed the advertising algorithm or Apple pushed new privacy policies, we saw a drop in performance and a scramble in the market to adapt. Not all companies managed to adapt, though. For some, each change like this was fatal.
Will access to AI become the “new wi-fi”? Something that is so ubiquitous, accessible, and affordable that we experience it more as infrastructure like running water than a high-tech service?
I think so.
However, I do not believe that GenAI is the final frontier of this evolution. I see “TRUST” as the most important currency of the future. If the information I consume on the internet is generated by AI agents known to have hallucinations and synthetic data, how do I form my perception of reality? Whom or what do I trust?
While access to information seems to be a commodity, the quality (trustworthiness) of information and the ease of getting it seem to be the new frontier of how we are evolving as a society.
Furthermore, even the existing market leaders are already thinking beyond large language models that underpin the current generation of Gen AI. Introducing a notion of planning into generation is one such promising approach. Personally I am also very keen on ideas around using knowledge graphs to better encode the real world and make reasoning less opaque.
While currently the majority of money flows into the lower layers (and is quite concentrated on just a few players), I could see that changing rapidly once we get more non-trivial functionality on the application layer. Currently large swaths of the application layer are just wrappers around ChatGPT, but what about once we start talking about data provenance, added value of data (and accompanying payment rails and microtransactions), data interchange, centers of trust…
And this is just the what. Also with the how I expect big and rapid movements on the market. The applications of AI to interfaces will be phenomenal. We are moving away from overdone interfaces and solutions are becoming more subtle. At the end of the day, we care more that the job gets done instead of who, what and how is doing it.
In terms of adoption, users will not try something just because you write “AI-first” as a headline on the landing page. Whenever you position something as AI-something, users will anticipate a 10x solution with very low time to value, usually provided for free and within usage limits. Do not put lipstick on a pig - think value-first.
We agreed that AI is important, here to stay and we should do something about it. But how?
How to start thinking about that?
Case 1: CommandBar: AI can help you fix your activation issue
My core principle is not reinventing the wheel whenever I need to optimize the existing process. Of course, you can build your own solution, but that needs to involve engineering and you won’t have the feature set of a product that’s been purpose-built for years.
If someone has already successfully solved a challenge unrelated to my core business and not within my product vision, I would definitely prefer to buy a tool that “hooks up something” and train it on my limited data samples. Usually, it is cheaper, better and faster.
When we talk about activation, it is often mistaken for onboarding—getting the user to experience the value of the product after the signup. In reality, activation starts much sooner than signup and continues beyond the first inception of the “wow moment.”
Emails, creepy support agents bots and annoying push notifications will be ignored. To truly make a difference within the product, modern-day SaaS providers and apps need to think way beyond that. Recently, I discovered a really cool visualization prepared by Kate Syuma, the author of Kate's Syuma Newsletter and Growthmates (ex-Miro Head of Growth Design):
CommandBar gets that. Instead of focusing on the onboarding experience alone, it offers powerful in-product engagement tools, intelligence and integrations to holistically tackle the activation challenge.
At its core, CommandBar is software that makes other software better for users. They call it a User Assistance Platform, and it comes in 2 pieces:
Nudge Platform — AI-assisted nudges, hints, and messages to guide users in personalized ways that don’t annoy them. They do this by optimally delivering nudges based on in-product behavior and not just static rules. They also measure the extent to which users “rage close” nudges to avoid pestering them.
AI Agent — an AI chat agent that can answer users' questions, perform actions for them, and even co-browse with them to teach the product’s interface.
As I dived into CommandBar and its references, I actually discovered that I have been using it for the last year or so without even knowing 😅
How come? We are using ConvertKit for our transactional and drip email campaigns, and I had a couple of very profound chats with their support chat. Little did I know it was an external tool because it is so seamless and beautifully branded.
I recorded a video overview of why CommandBar is one of the best customer activation tools available to SaaS companies. Remember: AI hallucinations can cost you reputation, time, and money.
As I wanted to dive deeper into this compelling product vision and vision of how user activation will evolve in the AI era, I invited James Evans to share more about the future of activation for the GTM Strategist podcast. One of my favorite parts of this episode was when he explained that they positioned CommandBar as a horizontal product - serving many industries for a similar use case.
Listen here:
You can capture the benefits of AI-assisted activation, get powerful purpose-built features that will improve activation and overall user experience and without investing 50-100K on your custom LMS solution development if that is not your core business.
But what if all the solutions to the problems you repeatedly face fall flat, and you see a window of opportunity? That might be considered developing yourself …
Case 2: $0 budget launch plan for Rewriter
Meet Tim Berce, my long-time friend and business partner. He is one of the best copywriters I have ever had the pleasure of working with. For years, we have been telling him that he should write more LinkedIn posts because he is a supremely talented writer, but that was simply not his spiel.
But he liked AI. Tim was one of my first colleagues who adopted OpenAI solutions before it was cool. So far, he has built several AI-first products, including:
Enterprise solution for writing perfect sales emails for a client
An app speeding up getting products to market for large retailers and marketplaces.
A smart email assistant
He needed to attract B2B clients for those tools, so he had to step up his LinkedIn game.
Tim is all about effectiveness and 80:20 rule.
Copywriting for LinkedIn is daunting, and as a busy venture builder, he cannot allocate 4 hours a day to writing about his cats and making valuable comments on other people’s posts.
He did what he knew best.
Tim trained the model on 60,000 viral LinkedIn posts and launched v01 of Rewriter.
First, he used it for his posts: https://www.linkedin.com/in/timberce. Tim scheduled a bunch of content generated with Rewriter in advance. Since he has been using the tool, he has gained many new followers and well-performing posts.
It worked for him.
Will it work for others too?
I love people who eat that dog food, so I immediately joined his private alpha.
V01 was a spreadsheet that looked a bit like this for repurposing hooks for my LinkedIn posts:
In the meantime, Tim put together a simple interface in Wordpress and so far, there are approximately 100 users in his beta. How did he get them? Easy-peasy:
He shared his tool freely with friends and asked them for feedback.
Tim did a quiet pre-launch on LinkedIn and gave people free access in exchange for their feedback.
All other users came via referral as his users tried out the tool and recommended it to others.
The best thing is that the users retained nicely (35%+), a fantastic early signal of product-market fit. They are also heavy users: they already created more than 4,000 LinkedIn posts.
I use Rewriter at least once a week when I have random ideas for LinkedIn posts (outside my content plan) that I want to test quickly. Here’s how it looks like:
A critical assumption we need to test next is who is the best ECP (early customer profile) for Rewriter. Based on the feedback from beta users, there are 4 hypotheses who can best benefit from the tool now:
Busy founders of small companies that want to write content on LinkedIn to get new leads
Copywriters and ghostwriters who work with multiple founders
Inhouse content creators that have to repurpose their content for LinkedIn
Influencers :D - professional content creators (15K+ followers) who spend more than 10 hours a week on LinkedIn and post at least 3-5x a week.
How will we know which is the best bet?
Easy.
We will launch and reverse-engineer data on who is most likely to activate and convert.
Tim and I had a launch planning session this week. The objective is to get from 500 (Tim’s bet) to 1000 (Maja’s bet) users, who will eventually become 20-50 paying customers.
Rewriter will launch in late August. To support the launch, we selected 3 free GTM motions.
Communities: Post to Facebook groups where target users are active.
Inbound + PLG: Brand posts created with Rewriter to create a growth loop
Outbound: LinkedIn DM & outreach to selected contacts.
Bonus points: Get 3 influencers to share it because they really like it. I volunteered to support this launch and one 105K+ LinkedIn influencer is also interested. Fingers crossed 🤞
It is simple, actionable and free.
If you like Rewriter, you can still get early access.
3 things you should not mess up when you launch an AI app
You see, Tim did not plan a Product Hunt launch. He will not present his side hustle at an event or spend money on ads to promote the tool. He will DIY this launch in probably a day. User referrals are already kicking in (this is a user referral, for example), so we are very bullish on the odds of this success.
How so? Because I have seen it working again and again.
One community member got 1000 users by posting to 3-5 local Facebook groups. I could not believe it. He showed me the stats.
If the value is there, users will follow and recommend it.
If the product is great and beta users love it to the extent that they already invited other users to join the program, we are likely to have a winner.
Most launches are completely overengineered, too expensive and unfocused.
You do not have to be a rocket scientist to get 1000 users for free - the best GTM for simple products such as Rewriter should be lean and actionable.
Use this template for your next launch if you agree with this line of thinking.
After we agree that launching an AI-first product does not have to be rocket science, let’s list some specifics you should consider for your own launches and offer more helpful sources.
Product: Beta or inception program is practically inevitable. At first, the product is likely to suck a little, and you need users who are passionate and patient enough to get it in a shape where it delivers value without your troubleshooting. Those users will also be your early ambassadors. You can get their reviews, and they will invite others to join. Definitely do not leave this one out.
Value prop/positioning: AI per se is not a differentiator. Probably the most brutal conversation I had this week was with someone who wanted to sell me 5 consulting hours. Look, I do not care about your consulting hours - you might be selling this, but I am not buying it. I care about how you can prepare me to nail my presentations in Q3. Name a price for that. It is harsh, but it is true. The same applies to AI - I do not care much about how this is done. What I care about is how I benefit from it, what are results I can expect from it and why you are the best solution to get this job done. That is it. FAB model (Feature, Advantages, Benefits) is a wonderful and simple tool that will help you get better in communicating your added value. Credits to Userpilot for sharing your example:
Pricing: AI usage has considerable costs - If you are developing an AI app, you will have real costs with “free users” beyond the marginal costs of web hosting services. Sure you can apply for inception programs for AI models and other players in the game to get free credits to build, but start thinking beyond that. You will need to reflect this cost in your pricing. Study this epic Substack post by Kyle Poyar so you will not get broke if the free users love your product too much.
There are many many more lessons that I have learned from working with dozens of AI products. It is my new favorite thing to do. My learning curve is exponential ATM so imagine how awesome my presentation at “What is the Go-to-Market Strategy for AI Products?” will be - apply here.
Cool, I hope to enjoyed this one.
If not - reply “Maja stop being so obsessed with AI” to this email or in the comments.
If yes - hit a heart ♥️ on Substack and send me more questions.
♥️ Credits to post contributors: The quality of this post is a result of fruitful long-term collaboration with:
AI part was fact-checked and improved by Simon Belak, data scientist and CTO (prev. Metabase, Zebra BI). He is one of the most intelligent people I have ever worked with. He never agreed that I put CTA next to his name, but I usually rely on him for complex analytics and data scient projects and technology leadership advice.
CommandBar team kindly pitched in with more insights on activation and their AI features.
Huge thanks and kudos to Tim Berce for having a true growth mindset and for openly sharing his data. #buildinpublic.
And to my team for all the awesome designs and editing this Substack. You are the best.
Let’s go to market with our AI-powered products!
Maja Voje
Are you ready to dive deeper and work on your Go-To-Market strategy and execution?
Discover my solutions that support your journey from launching to scaling:
🛰️ Explorer: Understand Go-To-Market and develop a winning GTM plan.
Get best selling GTM Strategist book + 20 frameworks (workshops) + online course ($47). Tested by 6500+ companies🚀 Doer: Leverage the 100-step GTM Checklist tested on 650+ launches with templates (emails, launch plans, posts, landing pages), which will guide you from getting ready to launch to an impactful launch and scaling stage.
Get the Checklist ($97)💫 Leader: Guide your team to successfully choose ICP (target market), pricing, positioning & selection of best GTM Motion (channels, tactics). GTM Bootcamp includes 6 hours of applicable videos from top experts.
Get the Bootcamp ($197)🏆 Ready to win: Tackle the most pressing Go-to-Market issues such as selection of ICP, positioning, how to get customers, build teams and select proven ways how to go to market. Book a 90-minute hands-on consultation with Maja, get personalized guidance, a Miro board, and reliable vendor contacts for further work.
Book here for $500
Curious :) what Facebook groups did you post on? I once considered FB, but the groups I found back then were too spammy