The consensus seems to be that AI can't do ‘taste’. As in, it can’t achieve that human-like level of discernment that separates the sublime from the subpar. I think that crowd is going to lose.
Taste isn't fixed. It's the accumulated reaction to whatever's in the culture at a given moment, and AI is a bigger part of the zeitgeist every week. People already consume AI-generated content constantly, often without realizing it. And the dirty secret is that a lot of it resonates. The line between "good taste" and "what people respond to" gets blurry when the thing they're responding to is already everywhere.
We've been heading here for a while. Hollywood doesn't make the 38th Marvel movie out of artistic conviction. It makes it because the formula tests well and plays to the widest possible audience. That's more replication than creativity, and replication is the thing AI does best. As the models improve, "AI can't do taste" is going to sound less like a principle and more like nostalgia.
But I also think there's at least one job I'd keep away from it anyway.
Last year I took over a forty-year-old consumer brand with around seventy thousand existing customers. Premium products, real loyalty, a name that already stood for something specific to the people who bought it. The marketing hadn't been touched in years, though, and it showed. So we rebranded: new logo, new site, new voice.
Here's the part that's easy to miss. We weren't working from an empty plot of land. We were renovating a house people already lived in. You can redo the kitchen. You can't move the front door upstairs.
Because a brand that's been around a while is a kind of shorthand. Mercedes means something — German, well-engineered, quietly expensive — and that shorthand is worth more than any single car they build. Stick that iconic logo on a Ford F-150 and you've broken the shorthand, even though the F-150 is the best pickup truck in the world. The badge stops meaning what it meant. That's the real risk in rebranding an established name: not that you'll make something bad, but that you'll inadvertently make something that no longer means what it used to.
AI can make something tasteful. Increasingly, it can make something good. What it can't do that a brilliant rebrand can is carry the shorthand forward, changing and modernizing the surface without cracking the foundation underneath. It doesn't know what it would be breaking, because it was never on the inside of the relationship. Taste is a moving average of the whole culture. But a brand's meaning is a private agreement with the people who already bought in, so the core essence can't drift.
In our case, you could see the difference in the work. The first logo concepts we got were… fine. Competent, but completely generic. They didn't break anything; they also didn't carry anything forward. Even though we were using an outside agency, the version that finally leapt off the page came from someone on our own team who'd been with the company far longer than I had. She knew the thread, so she could change the look without losing it. Same with the copy. We'd used AI for placeholder text while we were building out our new website. It, too, was ‘fine’, descriptive but not evocative. The copywriter we eventually hired wrote lines that worked because he'd listened for what the customer was actually looking to achieve and what they were afraid of, underneath what they said they wanted. We later fed AI models those same brand briefing documents and got flat results every time. The information was never the constraint. Carrying the meaning was.
This exception has limits worth naming. Run an express oil-change shop or a laundromat and there's no shorthand to protect. AI will take care of whatever branding you need well, fast, and cheap. Run a pre-revenue startup with no customers and AI's ability to throw a hundred variations at the wall is a gift; you have no continuity to break because you haven't built any, and can iterate more or less freely. The exception only bites when the brand already means something. The older and more specific that meaning, the more you need someone who carries it in their head.
Reducing everything to “AI versus human” is an oversimplification. The type of work matters, and down to the specific task. A blank page is a generation problem, and AI is great at those. Changing a forty-year-old brand is a continuity problem, and continuity is something you can only protect if you actually know it deeply. Know which one you're looking at, and you'll know when to keep your hands on the wheel.
Zain
P.S.: What does your brand stand for, in shorthand? And what's the one change that would quietly break it? Reply and let me know.