November 2026 · 4 min read
Taste: why it matters, and why AI can’t replace it
I work with AI, and I help businesses put it to work. It’s a fast pair of hands. But speed was never the hard part of design. The hard part is knowing which of the hundred things you could make is the right one, and being able to say why. That’s taste.
Decisions, not defaults
Taste isn’t a style, and it isn’t a set of rules. It’s a feel for what’s right: knowing something works before you can say why, and caring enough to get it there. In my experience it’s mostly something you have, not something you’re taught.
You can see it in a single card. Every default in a tool is a decision someone made for no one in particular: a size, a gap, a colour, a word. Taste is going through them one by one, and deciding on purpose.
None of those decisions is dramatic on its own. Together, one card feels considered and the other feels like a template. Most people couldn’t tell you why. They just trust one more than the other.
They aren’t rules, either. Another designer could make different decisions and be just as right. What matters is that each one was made on purpose, by someone who could tell you why.
Why it matters more now
When making something took weeks, the effort filtered out a lot of bad ideas before anyone saw them. Now you can have ten versions of a screen in a minute. Making has become cheap. Choosing hasn’t.
That moves the value. If anyone can generate a decent landing page, a decent landing page is worth very little to anyone trying to stand out. What’s worth something is knowing which of the ten is right for this product and these people, and what to change about it.
Being able to say why matters as much as being right. In a team, ‘it looks better’ doesn’t settle anything. ‘The number is what they came for, so it leads’ does.
What AI can’t do
A model gives you a likely answer. That’s what it’s built to do, and for plenty of work it’s exactly what you want. In design, the likely answer is the one everyone has already shipped.
Ask for a dashboard and you’ll get the dashboard: cards in a grid, a chart at the top left, the numbers in a row. It isn’t wrong. It’s the answer to the question everyone asks, and it looks like it.
That’s why so much AI design feels flat. It isn’t bad. It’s characterless and emotionless: no point of view, no feeling, nothing anyone meant. Taste is where the character and the feeling come from, and that’s what makes it powerful.
It also hasn’t been in the room. It hasn’t watched someone hesitate over a button, felt a flow go on one step too long, or noticed that the number everyone needs is the smallest thing on the screen. Taste is built from moments like those, and it belongs to someone who has to stand behind the result.
And it can’t care. Part of taste is saying no: to the extra colour, the third font, the animation that’s there because it could be. A model will happily give you all three. Someone has to decide, and answer for it.
Until a model can comprehend taste, or emotion, and who knows if it ever will, what it makes won’t quite have the feeling of design that was crafted on purpose.
A fast pair of hands
None of this makes AI less useful. I think it makes it more useful, in the right place. It’s very good at the parts of design that are really production: variations, first drafts, checks, the twentieth size of the same icon. It can explore wider and faster than I can alone, and it never tires of trying one more version.
What it can’t do is decide which of those is good. That’s why I see it as a tool that makes designers with taste even faster at ideation, not a replacement for them. I use it as a fast pair of hands, and keep the judgement for myself. It’s how I set AI up for the businesses I work with, too: it goes where it saves real time, and people stay where the decisions are. See how that works.
A fine line
I’m sure taste can be learned, by some. But I’ve worked with hundreds of designers, and I can say with confidence that it’s a fine line between those who can learn it and those who can’t.
That’s why I think AI will widen the gap, not close it. Designers who are creatively driven, have taste, and understand the difference between good and bad design will go further with it than they could without. Those who use it and let it answer their questions for them will fail, and be replaced.
AI can make the work. Knowing what’s good, and why, is still the job.