In short: As AI makes competent work cheap to produce, the scarce skill becomes taste: judgment about which option is actually right for a specific audience and purpose. Taste is built through exposure, consequence and apprenticeship, and organizations should hire and develop for it on purpose.
A few years ago, the hard part of most creative and strategic work was production. Writing the copy. Designing the screens. Building the model. Drafting the plan. If you could produce good work reliably, you were valuable, because production took skill and time and most people couldn't do it well.
Production is getting cheap. A capable model can produce a dozen headlines, four layout directions, three strategy options and a draft deck in the time it used to take to schedule the kickoff. Most of what it makes is competent. Some of it is good.
Which raises the question that now matters more than any other in this kind of work: which ones?
What Is Taste?
Taste gets talked about like a mysterious gift, something certain people are born with. I don't think that's right. Taste is judgment about quality, built from exposure and consequence. It's what happens when you've seen enough good work, made enough bad work and watched enough of both meet the real world that you can tell, quickly and often wordlessly, which is which.
It shows up as small, confident decisions. This headline, not that one. This idea is clever but wrong for the client. This design is clean but forgettable. This plan is thorough but nobody will follow it. None of those decisions can be proven in the moment. They're bets on what will work, made by someone who has been right often enough to be trusted.
Picture a homepage redesign. A model gives you twenty directions in an afternoon. Eighteen are competent. Two are interesting. One is right for this client, because it says something true about them that the others don't. Production got you twenty. Taste gets you the one.
Why Does AI Make Taste More Valuable?
You might think a tool that generates endless options would make taste less important. The opposite is true.
When options are scarce, the work is producing one good one. When options are abundant, the work is choosing. A designer who used to make two concepts now gets twenty. A strategist who used to write one plan now gets five. The bottleneck has moved from making to judging, and judging is exactly where taste lives.
There's a second effect too. Models drift toward the middle. Their output is fluent, reasonable and safe, the average of everything they've seen. Left to its defaults, AI produces work that is never embarrassing and rarely memorable. Taste is what notices that. It's the voice that says "this is fine, and fine isn't the goal."
Why Can't AI Do the Last Ten Percent?
I've written before about the last ten percent of any piece of work, the stretch where something goes from good enough to actually right. AI is extraordinary at the first ninety. The last ten is almost entirely taste: the word that's slightly off, the spacing that feels crowded, the argument that technically holds but doesn't land, the idea that's correct but not quite true to the client.
That stretch can't be generated because it isn't a production problem. It's a perception problem. You have to see what's wrong before you can fix it, and seeing is the skill.
Can Taste Be Taught?
I think so, though not quickly. Taste is built the way any judgment is built: through volume, feedback and attention.
See a lot of great work. Not just in your field. Taste borrows across disciplines. Architecture teaches software, editing teaches strategy, music teaches pacing.
Make a lot of work and watch what happens to it. Taste is calibrated by consequence. You learn what lands by seeing what lands.
Articulate your choices. When you pick one option over another, say why, even just to yourself. Taste that can be explained can be taught, questioned and improved.
Work next to people with better taste than yours. This is how most of us learned. A senior designer who winced at a kerning choice. An editor who cut the sentence we loved. Apprenticeship is still the fastest way taste moves from one person to another.
Isn't Taste Just Subjective?
Taste can sound like snobbery, a word people use to defend their own preferences. Sometimes it is. But good taste isn't liking the right things. It's knowing what will work for a specific audience and purpose, and being right about that often enough to be trusted. It's judged by outcomes, not by who claims to have it.
There's a second objection: models are getting better at critique, and they can be tuned toward what people prefer. They can already rank options against criteria you give them. That's useful, and worth using. But choosing the criteria, deciding what this client needs to feel and what this moment calls for, is still the bet. A model can score the options. Someone still has to own the choice.
What Does This Mean for Teams?
If taste is becoming the scarce skill, then a lot of hiring and development is pointed the wrong way. We still screen for production: portfolios of things people made, tests of what people can build. We should be screening for judgment: show someone ten options and ask which one and why.
And we should protect the people building taste. Junior team members develop it by producing work and getting honest reactions to it. If AI does all the producing, and seniors do all the choosing, the juniors never get the reps. We'd be spending down a skill we're not replenishing.
The Skill That Stays
Every tool that has made production easier has made judgment more valuable. The camera didn't end painting. It forced painters to decide what painting was for. Desktop publishing didn't end design. It made good design easier to spot against a flood of bad.
AI is the biggest version of that shift yet. It will make almost anything. It can't tell you whether what it made is any good, not in the way that counts, which is the way a client feels when they see it. That judgment is still ours. It might be the most important thing we have left to offer.
Frequently Asked Questions
What is taste in creative and strategic work? Judgment about quality: knowing which of several competent options will work for a specific audience and purpose, and being right about it often enough to be trusted.
Can AI have taste? AI can rank options against criteria you give it. Choosing the criteria, and owning the bet on what a client or moment needs, still takes a person.
How do you develop taste? See a lot of great work across disciplines, make a lot of work and watch how it lands, explain your choices out loud, and work next to people whose taste is better than yours.
How should teams hire for taste? Screen for judgment, not just production. Show candidates several options and ask which one they'd choose and why.
Evan Prout is a strategist in Grand Rapids, Michigan, who writes about how organizations think and decide with AI.