AI is part of how I work now, not a gimmick I bolted on. It's changed my process a lot more than it's changed my job. Here's what that looks like day to day.
AI is fast at the parts of design that were always a little mechanical for me: reshaping a messy doc, wiring up the tenth version of a flow, turning a decision into a screen I can actually click on. I use it for all of that, and it hands me back a lot of time.
What I spend that time on is the part I actually care about, and it hasn't changed at all: talking to users, working out what's worth building, and getting a team aligned around it.
The tools made me faster. They didn't start making the decisions.
Every project I take on starts as a repo instead of a Figma file: the research, the interview notes, the requirements, and the reasoning behind each decision all live in one place, and the AI is reading the same source material I am. When I start designing, I work against the real product, its actual components, tokens, and API contracts, so I find out where an idea falls apart early instead of after someone's written the ticket.
The prototypes that come out of that aren't throwaway. They run on real data, they break in the real places, and they're what I put in front of users. The good ones end up being what engineering builds from. It also means design, product, and engineering are all arguing about the same real thing, which is usually when the good decisions happen.
Honestly, most of my week doesn't involve any of these tools. It's people. I run design sprints and workshops, usually when a team's stuck or a room can't get to a decision, and I genuinely like the messy work of helping a group land somewhere they believe in.
I coach designers, which mostly looks like critique, pairing, and being honest with someone in a way that actually helps them. And I spend a lot of time on how design gets made on a team: where things keep getting stuck, which rituals earn their place, and what's just in the way. It's not glamorous, but it's usually what decides whether the rest of it turns out any good.
There are a few things I won't hand off to any of this, no matter how good it gets. Deciding what's genuinely worth building is a judgment call, and the tools will cheerfully hand me a hundred options without ever saying which one matters. Hearing what a user actually means, underneath what they literally said, is listening, and I've never found a way to outsource that.
Knowing which of five good-looking directions is the right one comes down to taste, and taste takes years. And getting a real group of people to agree is human work no prompt has done for me yet.
If every one of these tools vanished tomorrow I'd be slower, but the work would come out the same, because none of the parts that make it good were ever the tools in the first place.