Why We Still Pay Human Designers in 2026
We tested 6 AI design tools. Results surprised us.
Stepping up your inbound this summer? Ahrefs invited me to a format called Battle, and it’s exactly as fun as it sounds.
Next Friday (July 17), I’m going head-to-head with my very good friend (and future nemesis) Jairo Guerrero, co-founder of Organic Hackers - scarily good at technical SEO and at digging up the hidden keywords worth ranking for.
My weapons: MCPs, a Claude custom project, AEO, and the SEO workflows we built for B2B content. His: everything technical I can’t touch. Someone has to win, so come stand in my corner and help me whip votes - and either way, you leave with a fresh stack of ideas for getting more out of your content.
🎁 Bonuses: Jairo will share his client template Google Sheet for analyzing content gaps with everybody who attends, and I’ll update my AEO/SEO skill process on GitHub and share the link at the webinar ✌️
Dear GTM Strategist,
Design has always been my passion.
I’m kidding.
I have zero theoretical design knowledge, but I had to learn it on the fly. For two years, I wasted my weekends drawing shapes in Miro every Saturday and praying a human could decode whatever was in my head.
And yet design eventually became one of my biggest moats. English is my second language, so copywriting was never going to be the terrain where I could win globally within reasonable effort.
Design could be it. It carried a huge part of how GTM Strategist is positioned today, and it did more heavy lifting for our growth than I ever expected when I started.
So this is the piece where my husband and business partner Anže Voje and I open up our AI-assisted design process - the good, the bad, and the parts we still refuse to hand to a robot.
Big spoiler: I don’t think AI can do this job for us just yet. We still work with designers, on purpose, because branding is a moat and sometimes you need that final sparkle ✨, the 10% that makes it special.
My brother runs a men’s t-shirt ecom store doing serious volume, and he happily promotes it with AI-generated cartoon videos. For him, it works beautifully - because those creatives are meant to be broken - I mean, tested. If he has some winners from the batch, it is a win that pays huge bucks. It is almost a system in which we accidentally stumble into a winning creative you’d never have designed on purpose. Volume is the strategy.
Our visuals are the opposite. They’re built to (hopefully) last; we have a couple of shots a day to make or break it on LinkedIn or in a newsletter, and they are done with the intention of becoming learning assets people save and come back to.
That kind of work needs thoughtfulness, not thousands of iterations to test.
I would die of old age if I waited for legit test results from the velocity of publishing 1x a day on LinkedIn.
Some designs are made to be broken. Ours are made to last. That single distinction decides where AI belongs in your process - and where it doesn’t.
Our agenda for today:
Why brand is still a moat in 2026, and how the “made to last vs. made to be broken” line tells you exactly where AI fits
Our hybrid AI-human design system (big infographic), tool by tool, from idea to final file
Anže’s head-to-head test of six AI design tools on the exact same prompt. Do you wonder which one is the winner?
Let’s get into it.
Brand as a moat kind of just happened
When I published my book, I made a decision that looks obvious in hindsight: I was going to compete on brand and design. Everything in our industry was so boring - I wanted to make it more fun and memorable. If everyone is designing as if they were still making a management consulting presentation full-time, I just wanted to stand out by investing more in the brand and bringing it to life. Oftentimes the brand 1.0 is just an extension of the founder’s personality and I don’t think that is bad because you personally have to stand by every design choice. And it attracts people - people who share your values, joie de vivre, passions and fears.
The design process actually started with my cousin having a summer internship with us - a junior industrial and material design student, extremely creative and tech-savvy. I believed in him so much - but we were all just figuring out how to do this (and had a bunch of fun doing it). Before you roll your eyes: I did not hire him because he’s my cousin; we hire for attitude. We were genuinely a good combo.
Then something very predictable happened. The work evolved from print into LinkedIn graphics, and I learned the hard way: almost no brand designer enjoys tweaking social media posts. It is hard to find people who find joy in nudging a headline 4 pixels to the left on a carousel for the hundredth time. So when my cousin eventually graduated and moved on to pursue his true passion in life (industrial design), the visual heavy lifting landed where it usually lands - on Anže and me 🙈
We don’t see our visuals as pretty decoration or funny memes. They’re educational design. They’re meant to teach something and be worth saving. That’s why I stayed so hands-on with the structure and the microcopy, even though I was constantly feeling guilty about how much time and energy it consumed.
At one point, design was almost a part-time job on top of everything else. And more often than not, I still wasn’t fully happy with the output (damn perfectionist), because I couldn’t perfectly express in a briefing what I could see in my head. I felt so frustrated and trapped.
Then came AI … and I was overjoyed when it became good at design - sort of.
In the beginning, I genuinely hoped I could DIY designs with the help of AI - that I’d sit there chatting with my AIs and hoping for a miracle. Then I saw the actual inputs and the reliability, and reality hit. For video, honestly, AI is already better than what most of us would produce by hand. For infographics - see the third section 🙈
(If you think we are there and I’m just testing the wrong tools, tell me in the comments what I should be trying. Maybe you know something that I don’t. )
Inside Our Hybrid: Humans + AI Design System
Our system explained in one breath: AI research of best practices + our own winning patterns + a human editor and designer + a learning loop that feeds our top-performing designs back into the process so the next round starts smarter.
The inputs that feed it are:
Our brand book. We run two winning formats: a dark system for the more technical content, and our good old purple with yellow highlights. Fun fact on the yellow: I found it while buying a car, when a salesman highlighted his phone number on a business card with a yellow marker. Eureka. I’m pretty sure people have copied that combo since, but I won’t take credit - the car salesman invented it, not me.
Best external practices. We manually save LinkedIn posts that use shapes and formats that clearly perform well, and we use AI to spot the winning patterns we couldn’t catch by hand. Colleagues also just send us good visuals, which - do that, it’s the cheapest R&D there is.
Our stored role-model graphics. This has been a massive upgrade in the last few months. We never start a brief from scratch. Every brief now carries an input, a couple of role-model graphics that have worked before, then the mockup, then the video instruction, then the back-and-forth. (You can get access to the full archive of my graphics by purchasing my GTM Checklist.)
Proprietary models - new ideas that I still hand-draw on paper or Miro. This is the most time-consuming part and the intellectual moat. Before I hand the brief over to design, I need to make sure the content holds up.
I uploaded my brand assets to Brandfetch, which makes collaboration super easy.
Now the tooling, which is where people always ask me what I actually use:
I brief in Miro. I’m not great in Canva/Figma - I’m much better in Miro, so that’s where the thinking starts. Then I do mockups in Claude, often in Claude Code, and prompt it further to get to a version one. Claude Code is amazing for design mockups, and here’s the unglamorous reason why: it builds on your previous outputs instead of starting from scratch, so it wastes far fewer tokens and is much faster than doing this in a chat.
For example, if I want Claude in chat to update the mockup it delivered, it will start drafting again. Meanwhile, Claude Code will just change a few lines of code in the previous output.
Claude Design is also a useful feature that works similarly to Claude Code and is purpose-made for design work. But it’s mostly focused on web design and slides and is less suitable for complex infographics; that’s why it’s not the focus of this newsletter.
When the mockup is ready, I record a video interpretation of the brief for the designer; the designer comes back with a v1, and we elaborate together over Slack and Miro until the graphic lands where it should. We use Notion to keep track of the process and the editorial calendar.
Is it perfect? Of course not. It’s still pretty heavy on labor. But the split is clear, and it’s held up for us:
AI structures the information and roughs out the mockup. The final 10% - the part that makes it unmistakably ours - is still humans, in a room, bouncing ideas and resolving creative frustration together. That friction is what makes it special. At least for now.
I want to be fair here, because someone will bring it up if I don’t. Our colleague Alex Lindahl, GTM engineer at Clay, produces basically all his visuals with zero design cost and zero designers, using Claude skills exclusively. He described exactly how here. So it can be done, and done well. I’m just not ready to give up the intimacy of the process yet - I’d have a bit of an integrity problem publishing fully AI-generated designs today. Ask me again in six months.
The Head-to-Head: Six Tools, One Brief
Every time the models leapfrog each other, I get twitchy and ask Anže the same question: “Should we switch to [new hotness] for our design mockups now?” Anže, being the calm and calculated one, doesn’t answer on vibes and FOMO. He runs the test.
So we pressure-test our system periodically for one reason: to stay honest and make sure we’re not clinging to something the market has already moved past.
Here’s how we kept it fair. Same design brief, given to each tool.
The six tools Anže put head-to-head: we first started with the most common ones: ChatGPT, Gemini, Claude (Chat), and Claude Code.
Then we added two wildcards: Mistral Vibe (European AI hope) and Z ai (Chinese rival that recently topped some benchmarks).
Here’s the prompt we used:
Draft me the mockup for a visual that will illustrate this LinkedIn post:
5 modes of LinkedIn for GTM, from a sharp profile to a full system. Pick the one that fits your goals and bandwidth right now:
Mode 1 - 𝐈𝐧𝐝𝐞𝐱𝐞𝐝 (The Foundation)
→ A profile that does its job the second someone Googles you or asks ChatGPT/Claude. Findable, credible, sharp. That's all you need if you're not running an inbound motion.
Mode 2 - 𝐎𝐩𝐢𝐧𝐢𝐨𝐧 𝐋𝐞𝐚𝐝𝐞𝐫 (The Engine)
→ A real POV plus the discipline to show up with it consistently. That builds trust and preference at scale - even if you only comment. Ideally, you do both.
Mode 3 - 𝐆𝐓𝐌 𝐄𝐧𝐠𝐢𝐧𝐞𝐞𝐫 (The Converter)
→ Skip the content purgatory. Focus on signals - what your ICP complains about, which events they attend, how they change jobs. Build a signal system and turn it into outreach that actually lands.
Mode 4 - 𝐇𝐨𝐮𝐬𝐞 𝐨𝐟 𝐏𝐞𝐫𝐬𝐨𝐧𝐚𝐥 𝐁𝐫𝐚𝐧𝐝𝐬 (The Multiplier)
→ Several profiles amplify one core story. Team lift, huge dividends - but a long game, and turning your best experts into "content creators" carries real opportunity cost. Worth it for many. Not for everyone.
Mode 5 - 𝐌𝐞𝐝𝐢𝐚 𝐌𝐚𝐜𝐡𝐢𝐧𝐞 (The Insurance)
→ LinkedIn can be a pain in the arse the second the algo shifts or they go on another automation witch hunt. One flag and you could lose a valuable asset overnight. Hedge it - build on channels you actually own (email), or diversify across social. Stop depending on one platform.I was shocked to get such a variety of outputs with a single prompt. 🙈 Mistral, what on Earth did you do?! Nano Banana, 🍌 in 2025 we called you superior …
ChatGPT is the undisputed heavyweight for visual generation right now, mainly because it isn’t just spitballing pixels. It blends the creative with solid contextual research and microcopy. The result was a sophisticated infographic that, while perhaps too dense for a quick scroll, is objectively impressive in its depth.
Claude Code (running Sonnet 5 as a fast and versatile model) had an unfair edge. Despite my specific nudge to keep the layout neutral, it reached into our brand context and guidelines anyway. It handed out a mockup that was already one step closer to the final design.
That’s why there was a striking difference with Claude in chat, starting from scratch (also running Sonnet 5). The chat version was disciplined, following the brief to the letter without any extra flair. It delivered a clean, basic outline that synthesized the core points nicely. Which is actually not bad, because the best educational visuals don’t drown in text.
Gemini and its Nano Banana image generation model might be flashy for pure art, but for infographics, it over-engineered the solution. It was the most “creative” output in the batch, but it strayed so far into its own imagination that it became nearly useless as a functional brief for our system.
Mistral Vibe was the letdown of the group. It just put the information into a basic diagram with some summaries and called it a day. No real spark there.
On the other hand, Z ai (running GLM-4.7) was a pleasant surprise. It was my first time using it, and it did a solid job of both distilling the knowledge and proposing a coherent layout and color scheme. Still, a word of caution here: Chinese AI chatbots have significant data privacy vulnerabilities, so I would avoid putting sensitive or personal information in them.
Both Vibe and Z used HTML for their designs, much like Claude Code. This is a massive workflow win: we can just hand that code to a designer, ensuring the copy stays intact and nothing gets lost in the process.
Verdict? For our regular work, we will stick with Claude Code, because it’s so well connected with our business context, skills, and workflows. But when we need additional creative ideas, ChatGPT is the way to go.
So, Should You Go Full AI?
Here’s where I’ve landed, and I hold it loosely because this space moves fast.
If your creatives are meant to be broken - high volume, disposable, find-the-winner - lean into AI hardcore. Your business probably depends on it because everyone is using it atm. My brother is right, and AI-generated content helps him experiment much faster at a marginal cost, while he still combines it with three other content pillars: static photos, user-generated content, and high production.
If your creatives are meant to last - brand assets, educational design, the stuff people save - keep humans in the loop for inception and at least that final 10% of design touch - often referred to as “taste”. Use AI where it’s genuinely great: research, structure, and early mockups.
And here’s the timely part: we’re actually hiring a designer. Our production velocity is too limited right now, and Anže and I have taken this system about as far as the two of us can. We’d love someone to elevate it further.
It’s a global, remote role, part-time or full-time - we left it open on purpose so we don’t scare off great people over hours. The one real requirement: proven design work and content work for LinkedIn, newsletters, and blog posts - the good old B2B marketing stuff. If you vibe with this philosophy, apply here.
If you want to level up your design with AI, I encourage you to spend a little of this summer structuring your mood boards, finding your winning patterns, and tweaking your design system to play in your favor. That’s the part AI can’t shortcut for you.
Let’s make LinkedIn a little less nebulous - fewer random patterns and soulless gradients, a bit more human spark ✨ - so we can actually learn from each other and grow together.
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.
Let's go, I'm available and curious. Actually did a round of test tasks in this scope and niche with you both earlier this year, and would love the chance to work together again now that you're growing this out. Applying today!! Thank you :)