Everyone keeps telling designers the craft is dying. They’re looking at the wrong thing.

A few months ago I sat in a review where a designer presented forty screens in under ten minutes. Clean. Consistent. Shipped-looking. The kind of output that would have taken a small team a week not long ago, done overnight by one person and a model.

Someone on the call said what everyone was thinking. “This is incredible. We barely need designers for this anymore.”

I let it sit for a second. Then I asked one question. “Which of these forty are we actually going to build — and why those?”

The room went quiet. Nobody had an answer. Forty screens, and not one person could tell me which one was right. That silence is the whole story of craft in the AI era, and it is the opposite of the story everyone is telling.

First, the part the doomsayers get right

Let me give the pessimists their due, because they are not entirely wrong. AI genuinely did kill a large part of what we used to call craft. The part that was labor.

For twenty-five years, a real share of a designer’s day went to production. Pushing pixels into alignment. Resizing the same asset eleven times. Rebuilding a layout because the copy changed. Redlining specs by hand. That work ate hours and rewarded patience more than talent. If your definition of craft was “the hours of careful execution between the idea and the finished file,” then yes — that craft is dying, and quickly.

I am not going to romanticize it. I did that work for years. Most of it was not the reason I got into design, and most of it was not where I added anything only I could add. When a model does it in seconds now, I do not mourn it. Good riddance.

Now the part they get wrong

Here is where the argument falls apart. The people declaring craft dead are making a quiet substitution. They are treating execution and craft as the same word. They were never the same word.

Execution is the labor. Craft is the judgment that governs it. It is the thousand small decisions that a model, by its nature, averages away.

Craft is the kerning nobody notices until it’s wrong. The one word in the headline that lands, instead of the three that don’t. Knowing when a thing is finished, and having the discipline to stop before it’s over-designed.

A model is trained to produce the most probable output. Craft is very often the improbable one — the choice that is right precisely because it is not what everyone else would have done. AI is extraordinary at giving you the average of everything that has worked before. It cannot tell you which of its forty confident options is the one worth your company’s name. That has never been a labor problem. It has always been a taste problem, and taste did not get automated. It got more valuable.

Why this matters more now, not less

Play it forward. When execution was expensive, competence was a moat. Being able to reliably ship clean, professional work was itself a differentiator, because it was hard and slow and most people could not do it well.

That moat is draining as we speak. The feed, the pitch, the portfolio pile on every hiring manager’s desk — all of it is about to fill with work that is competent, AI-assisted, and technically fine. Fine becomes the floor. Everyone is standing on it.

And when everyone can produce fine, fine stops meaning anything. The only thing that rises above a wall of fine is the work somebody actually cared about. The work where a person made the hard calls, sweated the details the model would have smoothed over, and took responsibility for the direction. Craft stops being the thing that slows you down. It becomes the only thing that sets you apart.

What this means if you lead a team

This is not just a craftsperson’s concern. It is a leadership one, and it changes what you should be hiring and rewarding for.

If you staff and measure your team around output — how much they ship, how fast — you are optimizing for the exact thing that just became free. You will win a race that no longer has a prize. The teams that pull ahead will be the ones that treat judgment as the scarce resource it now is: people who can look at a wall of plausible options and know, and defend, which one is right.

Back to that review with the forty screens. The designer who generated them was talented — genuinely. But the person the room needed in that moment was not the one who could make forty. It was the one who could kill thirty-nine. In an AI era, that is the whole job. Not the loudest person in the room. The most insightful.

So no, AI did not kill craft. It stripped away the labor that was hiding it and left the real thing exposed, where it should have been all along.
AI does fast. Craft does unforgettable. Craft was never decoration. It is direction — applied a thousand times over.

Joe Cahill has spent 25+ years leading design across enterprise and fintech. He writes about design, judgment, and leadership in the AI era.

Related Articles