A tool can now generate a hundred screens before lunch. Mockups, copy, even passable component libraries, produced faster than any of us can sketch them. The instinct is to panic or to celebrate. I think both miss the point.

Generation was never the bottleneck. The hard part of product design has always been knowing which screen to make, why, and what to cut. AI made the cheap part cheaper. It did almost nothing to the expensive part, judgment.

What AI actually changed for me

I've folded AI into nearly every stage of my workflow, and the honest answer is that it changed my speed, not my standards. I move faster through the parts that were always mechanical, which buys me more time for the parts that were always hard.

  • Discovery. I use AI to summarize research, cluster interview notes, and pressure-test my own assumptions, but I still talk to real users. The model can tell me what's common; it can't tell me what matters.
  • Exploration. Generating ten directions in minutes is genuinely useful, as a way to find the two worth refining, not as a way to skip refinement.
  • Production. Boilerplate, states, and edge-case variations are faster now. That's time I redirect into the details users actually feel.
AI is a fast intern with infinite patience and zero accountability. The accountability is still mine.

Where the craft lives now

If a model can produce the obvious version of anything, then the value moves to everything that isn't obvious. The judgment to kill a beautiful idea because the data says no. The systems thinking to make one decision scale across a product. The taste to know when "good enough" is a lie.

1. Problem framing

The quality of any AI output is capped by the quality of the question. Framing the real problem, not the one the brief states, is still entirely human, and it's where most of the leverage hides.

2. Systems over screens

A generated screen is a dead end. A token, a component, a pattern that holds across a hundred screens is leverage. I still build the system first, because the system is the thing AI is worst at and teams need most.

3. Evidence over opinion

AI is confident about everything, which makes it a dangerous source of truth. I keep validating with usability and A/B testing, because a model's guess and a user's behavior are not the same thing, and only one of them ships.

The takeaway

Treat AI as an accelerant for the work you already know how to do well, and a liability for the work you don't. The designers who win this era won't be the ones who generate the most. They'll be the ones who still know what's worth keeping.

None of this is anti-AI. I'd be slower and worse without it. But the tool raised the floor, not the ceiling, and the distance between them is exactly where a designer earns their seat.