AI Didn't Replace Designers: It Replaced the First Draft

A short one, because the realization is short.

I’ve been leaning hard on AI tooling in my day-to-day this fall: v0 for UI scaffolds, Copilot in my editor when I’m poking at prototypes, Figma AI for the little fill-in-the-blank tasks that used to eat an afternoon. The thing I keep noticing is not that AI is going to replace me. It’s that AI has already replaced the part of my job I was worst at: starting.

01. It killed the blank canvas

Here’s a recent example. I needed a settings page for a product feature, nothing exotic. In the old workflow I’d have stared at Figma for 20 minutes, made three rectangles, hated them, moved on to something else, and come back that afternoon to make progress. You know the dance.

Instead I typed this into v0:

Generate a responsive SaaS settings page with a sidebar navigation, a user profile form, and a red danger zone for account deletion.

Fifteen seconds later, it handed me a working React component. Not a great one. Not one I’d ship. But a working one. And the 20 minutes of stare-at-the-canvas paralysis was just, not there anymore. There was something on the screen to react to, and reacting is a mode I’m much better at than originating.

02. The taste gap

The catch, and every AI workflow post has a catch, is that AI generates aggressively average UI.

That’s a compliment, in a way. “Average” means it’s read enough examples to know what a settings page looks like. It puts the sidebar where sidebars go. It uses sensible spacing. It doesn’t invent a new interaction pattern. If average is what you need, you’re done in a minute.

Average is not what I need. Average is where the work starts. The first draft v0 gave me had subtle problems all over it: the danger zone was visually too similar to the adjacent section, the form labels were above the inputs in one place and beside them in another, the responsive breakpoints fired at weird widths. None of those are model failures. They’re just the gap between “a settings page” and “our settings page,” and that gap is where human taste earns its keep.

I’ve started to think of it as a taste gap. AI gives you the thing that satisfies the average of its training data, and your job is to drag it toward the specific thing your product actually needs. The gap is wider than I expected. The work inside the gap is more interesting than I expected.

03. What I spend time on now

My iteration speed went up. I don’t think that’s the interesting change. The interesting change is what I do with the saved time.

More accessibility checks. More time cross-referencing generated output against our design system tokens, because AI will happily use a hex value that looks right but isn’t in our palette. More time on edge cases and empty states, which is exactly the shift I should have made on my own five years ago and never quite did.

The generated code is a first draft. I’m the editor. Editing is a skill I’ve spent my career building, and it turns out to be the skill that survives.

04. Closing

If you’re nervous about AI in design, my advice, for whatever it’s worth, is to use it for the thing you’re worst at. For me that’s starting. You might find, like I did, that it doesn’t replace you, it just quietly eliminates the friction you’d built a career working around.

The floor came up. The ceiling didn’t move. What’s left is the climb.