How design work is changing
When AI makes the volume, design craft moves to judgment. What stays human, what I now hire for, and the risk I watch for.
When AI makes the volume, the craft of design moves to judgment. Making a screen, a component or an image is cheap now. Deciding which one is right, and saying no to the rest, is not. That decision is the work.
Making is cheap; deciding isn't
On LOU, my own game, Claude Code agents write the code, and AI tools generate the portraits, the voices and the music. My work is the bar: when the first mood portraits came back off-style, I rejected the model and paid more to make them all again.
On my PLL prototype, the agents built the full-width header banner the brief asked for, and it felt like an ad, so I replaced it. The agents did what they were told. The result still wasn't right, and only a person could see that.
Four things stay human
Talking to users. A prototype costs almost nothing now, so it's tempting to skip people. That's the wrong way round. At Fyndiq we tested ideas with prototypes before writing code. For the Insights section I designed at work, I built the prototype so research could test variants without a new build. Cheap prototypes should mean more research, not less.
Taste and the quality bar. Someone has to know when a thing is good enough, and when it's lovable. A model can produce a hundred options. It can't tell you which one your users will trust.
Trade-offs and decisions. Every real choice has a cost, and someone has to own it. At Fyndiq, landscape product cards scored better for scanning, but younger shoppers preferred portrait. We chose portrait and accepted the weaker scores. A tool can show you the data. Owning the cost is a person's job.
Growing people. I've led and coached design teams of up to 26 people. Trust, feedback and growth happen between people, and no tool does that part.
What I hire for now
I hire for judgment and for writing. A good brief, a clear decision note and an honest critique matter more than speed in a tool, because they are how a designer directs both people and AI.
I also want designers who pair with engineering. At PLL, I earned the engineers' trust first, and that trust grew a small library project into a redesign. Today a designer can work in code next to engineers. That means less handoff and more shared ownership of what ships.
The risk: everything starts to look the same
What worries me is sameness. Generated work drifts toward the average: the same layouts, the same illustrations, the same voice. If every team uses the same tools with the same defaults, every product starts to look alike.
Design leaders have to protect a point of view. Write the principles down. Set the quality bar and hold it. Say no to output that is fine but forgettable. That's not new. What's new is how much fine output there is to say no to.
That's the craft now: judgment.