Open LinkedIn any Tuesday and you’ll see another AI-generated rebrand of a heritage brand. “Coca-Cola if it launched in 2026.” Gradient mesh, friendly geometric sans, a mark that could belong to a fintech, a meditation app, or a cat food subscription. Very polished, completely interchangeable. Same story with the AI posters, the cartoon illustrations, the scratchy slanted fonts that were everywhere by March and invisible by June.
Thousands of people, the same tools, working independently, all landing in the same place. The tool isn’t the thing that failed. Taste is.
What changed when execution became free
Making things used to be hard and slow. Clients paid agencies partly for thinking, but mostly because agencies could produce. The kerning. The guidelines document.
The deck that didn’t look rushed at 5pm. Execution was the product, and thinking came bundled in for free.
AI hasn’t eradicated that. It has inverted it. When anyone can generate forty competent options in an afternoon, the value of those options drops to nothing, and the value of knowing which one, and why, becomes the whole game.
The cost of making collapsed, so all the value pooled in choosing. Taste stopped being an aesthetic preference and became a decision-making discipline, the difference between forty things that could exist and the one that should.
You can already see the gap
Surreal sells cereal off the back of ads that are, executionally, almost nothing. Flat colour, one line of copy, a joke about hiring random men called Dwayne Johnson so they could technically claim a celebrity endorsement.
Plenty of people could have made those ads. Not many would have run them. The judgement, knowing exactly how far to push the bit and that the category was earnest enough to crack under one good joke, is the entire brand.
No model arrives there, because models optimise toward the plausible, and the whole joke lives in the implausible.
Now the other direction. Last December, McDonald’s shipped a fully AI-generated Christmas ad in the Netherlands and pulled it three days later under an avalanche of mockery. Ten specialists, seven weeks, thousands of generated takes.
Enormous effort went into the making. What nobody in that pipeline did was stop and ask whether it should go out at all, even though Coca-Cola had taken the exact same beating the Christmas before. Near-unlimited capacity to make, and the entire failure lived in a single decision nobody exercised.
Luxury fashion ran the trial run for this a decade early. Through the late 2010s, the great houses filed their serifs down into the same flat geometric sans, each decision defensible on its own,the collective result being century-old brands converging on a default setting.
Then Burberry reinstated its serif and the Equestrian Knight, deciding a hundred and twenty years of accumulated character was worth more than fitting in. Serifs, crests, ornament, all the texture the minimalist years had sanded off, became the point of difference precisely because nobody else could claim it.
AI is that convergence times a thousand, arriving across every medium at once. The brands that escape it won’t out-produce anyone. They’ll out-decide them.
Taste is not what most people think
Here’s the part that gets romanticised. Taste isn’t a gift you’re born with. It’s pattern recognition plus conviction: having seen enough good and bad work, closely enough, to know the difference before the market tells you, and being willing to act on it when that’s inconvenient.
Ira Glass described the mechanism years ago. Your taste develops faster than your ability, and the gap between what you can recognise and what you can make is what kills most people early, before they’re good enough to satisfy their own standards. AI just deleted the second half of that equation.
The ability gap is gone. All that’s left is the recognising, which means the thing Glass called the painful part of being a beginner is now the entire job.
How you actually build it
Taste is learnable. The method is boring and unglamorous, which is exactly why it’s defensible. Nobody can shortcut their way to it, which means the people who put the reps in stay ahead.
Articulate, don’t just look. Pull up a hundred identity systems and write down why three of them work. Then debate it with someone who might disagree. The writing and the argument are the reps, not the scrolling.
Make the call before you know the answer. Commit to a direction in writing, date it, and go back in six months to mark your own work. Taste compounds through outcomes attached to your name, not opinions you never tested.
Get specific about what’s wrong. “I don’t love it” won’t help you. “The serif is doing heritage, the palette is doing startup, and the brand can’t be both” will. Vague reactions stay vague forever.
Defend a position longer than is comfortable. Taste without conviction is just curation, and curation is something models are getting good at.
Study the failures, not the award books. Everyone can explain why the winners worked, usually after the fact. Working out why a technically excellent piece of work still landed badly is where the actual judgement lives.
None of this is fast, and that’s why it stays human. The tools will keep getting better at making things. None of that improvement touches the question of what should exist in the first place, and that question was always the interesting one. It just used to be possible to hide from it by staying busy.
Pony Studio is the Emerging-Tech Brand Studio — a London-based branding and creative design agency specialising in strategic brand development for tech companies worldwide. If you’re building something bold and want a brand that moves at the same speed as your ambition, let’s talk.