AI Can Build the Screen. Your Job Is Knowing Which One.

A designer can now build three versions of a new onboarding flow in one afternoon. Two years ago that was a week of work. All three will look good. At least one of them will be wrong.

Not wrong in a pixel sense. Wrong in the sense that it asks a truck driver for a fuel card number while he's standing at the pump in the rain, with one bar of signal and a line of trucks behind him. The tool doesn't know that. A newer designer might not know it either. That gap is where our jobs live now.

I've led product and UX teams for a long time, and I have never seen the ground move this fast. Prototypes that took days take minutes. Engineers ship UI without waiting on design. Executives ask, sometimes out loud, why they need a design team when a model can produce a screen. If you work in product or UX, you've felt some version of that pressure. This is how I'm thinking about it, and what I'm asking my own teams to do.

The Work Didn't Disappear. It Moved.

AI has made output cheap. Screens, flows, copy, even working front-end code. When something gets cheap, the value shifts to whatever is still scarce.

What's scarce now is judgment. Knowing which problem is worth solving. Knowing which of those three onboarding flows a real person will actually finish. Knowing what to do when the dashboard says one thing and the customer sitting across from you says another.

If your value is "I make the screens," you're competing with a tool that doesn't sleep. If your value is "I know which screen moves the business, and I can show you why," you are more needed than you were a year ago.

Get Closer to Customers, Not Further Away

The trap I see most often is this one. AI makes it easy to produce research summaries, personas and journey maps without talking to anyone. They look great in a deck. They're also usually average, because they're built from average data.

The teams doing well right now do the opposite. They let AI clear out the busywork so they can spend more hours with real users. When we took our driver document workflow from ten steps down to three, that didn't come from a clever prompt. It came from watching drivers scan paperwork in a cab with bad lighting and worse signal, and seeing exactly where they gave up.

Use the tools to synthesize faster. Don't use them as a substitute for being in the room.

Learn to Build, at Least a Little

The line between design and development is thinner than it has ever been. I don't think every designer needs to become an engineer. I do think every product and UX person should be able to put a working prototype, connected to real data, in front of a customer.

This year I built a full back-office platform for a growing home services business. It started as a single page that priced a honey-do list. Forty-nine versions later, it prices jobs on site from a typed list or a photo, gets the customer's approval on their phone, and carries that same record through invoicing, scheduling, materials, receipts and payment. It runs on the company's own cloud for about a dollar a month. I wrote the specs, designed every screen, made the product calls and tested each release with the crew. AI was my engineering team.

A few years ago, software like that needed a funded team and a long runway. This time the hard parts were the human ones: turning a paper price sheet into something a crew could trust in a customer's driveway, deciding where AI could help and where a person had to stay in control, and making every screen work on a phone, out in the field, with one hand.

That's the mix worth growing. Enough technical fluency to make an idea real, and enough design sense to make it worth using.

Fix Your Design System Before You Scale AI

AI tools are only as good as what you give them. If your components are a mess, AI will happily produce more mess, just faster.

Our design system grew to around 200 reusable components across four products, and that foundation is a big reason AI helps us instead of creating chaos. When the building blocks are clean, documented and accessible, generated work starts closer to done and needs less cleanup. If you're looking for a project leadership will understand and fund, this is a good one.

Talk in Outcomes

This one is for anyone who wants a real seat at the table. Nobody funds "better UX." They fund lower support costs, higher conversion, faster onboarding and fewer customers walking away.

When we rebuilt our knowledge base and self-service experience, the story that landed with executives wasn't about the interface. It was that most customers could now solve problems on their own, and the company saved real money because of it. Learn the numbers your leaders already watch, and tie your work to them every time.

AI makes this easier too. Pulling analytics, sizing an opportunity and drafting a business case used to eat a week. Now it can take an afternoon, so there's no reason to walk into that meeting without numbers.

Hold a Higher Bar Than the Tool

Models are confident. They'll give you an answer that sounds right whether it is or not. Accessibility, edge cases, tone of voice, the moment a frustrated customer hits an error at 9 p.m. Those still need a person who cares how it turns out.

On my teams, AI output is a first draft and never the final one. Someone owns every decision that ships, and "the tool suggested it" doesn't count as a reason.

What I'm Telling My Team

If you're unsure where you fit right now, this is where I'd put my energy over the next year:

• Spend more time with customers, and let AI handle the note-taking and the first pass at synthesis.

• Learn enough code, or enough no-code, to put a working idea in front of people.

• Make your design system something both people and tools can rely on.

• Connect your work to a number the business already tracks.

• Own the judgment calls. Review what the tools produce like your name is on it, because it is.

The people who will struggle are the ones waiting for things to calm down. I don't think they will. The ones who do well will treat AI like a very fast, very literal teammate, and spend the time it gives back on the parts of the job that always mattered most.

Our tools changed. The job didn't. Understand people, then build things that make their day a little easier.