What AI Still Can't Do in Brand Work (And Why That's Your Moat)

The Question Founders Keep Asking Me

Every founder conversation I’ve had this year eventually arrives at the same nervous question, usually phrased as a joke. “Do I even need a brand agency anymore, or can I just prompt my way there?”

It’s a fair thing to wonder. The tools are genuinely good now. You can generate two hundred logo directions before your coffee gets cold, spin up a full landing page in an afternoon, and rewrite your positioning in the voice of six different companies you admire. The production ceiling collapsed. Things that used to take a design team a week now take an hour.

So here’s the honest answer I give them. AI has made the making of brand assets almost free. It has done close to nothing for the thinking that makes those assets worth anything. And the gap between those two things is exactly where your advantage lives, if you know how to protect it.

What AI Is Genuinely Great At

Let me be clear that I am not one of those creative people waving my arms about the machines. I use these tools every day and they’ve changed how our studio works. Credit where it’s due.

AI is excellent at volume. When we’re exploring a visual territory, generating forty variations of a concept surfaces directions we’d never have sketched by hand, and it does it in minutes. It’s a fantastic sparring partner for early-stage ideation, the kind of rapid back-and-forth that used to eat whole mornings. It handles the tedious middle of production beautifully: resizing assets across fifteen formats, drafting alt text, cleaning up a messy transcript, turning one blog post into a week of social copy.

It’s also a great leveler for the parts of the work that genuinely should be commoditized. Nobody should be paying premium creative rates to have someone manually kern a hundred social tiles. That’s a solved problem now, and freeing humans from it is a gift.

For a lot of tech startup branding work at the execution layer, AI is a real force multiplier. If your bottleneck is production speed, the bottleneck is gone.

Where It Quietly Falls Apart

Here’s the part that doesn’t show up in the demo videos.

AI is trained on what already exists. That’s the whole mechanism. It predicts the most probable next thing based on everything it has absorbed, which means its default output gravitates toward the statistical center of whatever you asked for. Ask it for a “clean, modern SaaS brand” and it will hand you the clean modern SaaS brand, the one that lives in the shared imagination of ten thousand training examples. It’s competent. It’s also the exact thing forty other startups in your category are generating from the same prompt this week.

That’s the trap. The tool that makes you fast is the same tool making everyone else fast in identical ways. When production is free and everyone draws from the same well, the output converges. I’ve watched entire funding cohorts arrive at visually indistinguishable brands, each one convinced they’d found something distinctive, none of them realizing they’d all typed roughly the same prompt into roughly the same model.

Strategy is where it really shows. AI can describe your positioning back to you in polished language, and founders mistake that fluency for insight. But real positioning comes from a decision, a genuine choice about who you’re for and who you’re willing to lose. That decision requires taste, conviction, and a tolerance for being disliked by part of the market. A model optimized to please cannot make that call for you. It will always nudge you toward the version everyone can nod along to, and a brand everyone nods along to is a brand nobody remembers.

Then there’s context. A model doesn’t know that your biggest competitor just repositioned last month, that your CEO has a specific way of talking that customers love, that your category has an unspoken cliché everyone’s tired of. It doesn’t feel the room. Brand work is mostly judgment about a particular company at a particular moment, and judgment is precisely the thing that doesn’t generalize.

The Moat Is the Part You Can’t Prompt

So where does that leave a founder trying to build something that lasts?

The uncomfortable truth is that AI has raised the floor and done nothing for the ceiling. Everyone can now produce competent brand assets. Competent is the new baseline, and a baseline that everyone shares is worth almost nothing as a differentiator. The value has migrated entirely to the layer above execution, the layer AI can’t reach.

That layer is judgment, point of view, and the specific human decisions that make a brand feel like it came from someone rather than something. A distinctive strategic position. A verbal identity that sounds like an actual person made a choice. A visual system with an idea inside it, not just a competent arrangement of the current conventions. These are the things a good creative agency for tech companies actually sells, and they’ve become more valuable now, not less, precisely because the surrounding production got cheap.

Look at the AI startup branding examples that genuinely stand out right now. Every one of them used AI to move faster on execution, and every one of them had a sharp human decision at the core that no model would have volunteered. The tool built the assets. A person supplied the conviction.

How to Actually Split the Work

Here’s the practical version I give founders, because the abstraction only helps so much.

Automate the production. Let AI handle volume, variation, resizing, first drafts, and the repetitive middle of the work. Push it hard here and don’t feel precious about it. Speed is a real advantage and you should take it.

Protect the decisions. Your positioning, your point of view, the specific tone of your voice, the one idea your visual system is built around, these are the things you make slowly and on purpose, ideally with someone whose taste you trust. Do not let a probability engine make them by default, which is exactly what happens when you accept its first fluent-sounding answer without a fight.

The test is simple. If the output would be roughly the same for any company in your category, automate it and move on. If the output should be unmistakably yours, that’s human work, and it’s the work worth paying for.

The Founders Who Win

The startups that come out of this era with strong brands won’t be the ones who resisted AI or the ones who surrendered every decision to it. They’ll be the founders who understood the split. They’ll use the tools to erase the grunt work and reclaim their time, then spend that reclaimed time on the handful of decisions that actually determine whether their brand means anything.

AI gave everyone the same superpowers at once. When everyone can produce, production stops being the game. What’s left is taste and judgment and the nerve to commit to a point of view, and those were always the hard part anyway. The machines just made it obvious.