AI is a paintbrush. You still need a painter.

The problem in 2026 isn’t that AI can paint. It’s that everybody now thinks they can.

You sit down with one of these tools, type what you want, and a minute later there’s a product on the screen. It’s got a nav, a colour palette, a pricing page. It looks like the thing somebody was going to charge you a few grand for. So the question comes, reasonably: why would you hire a designer?

This week’s video answers it with five reasons, and none of them are the ones you’d expect from a designer trying to protect his job. This is the fifth, and it’s the one I’d defend hardest.

Subscribe on YouTube (opens in a new tab)

Let me be straight about the counter-case first

A bloke built Base44 on his own, no design team, and sold it to Wix for around eighty million dollars, reportedly about six months in. Lovable’s entire pitch is that people who aren’t designers can ship software, and it’s running at serious revenue. I’m not going to pretend that isn’t happening.

I’m not going to pretend the design job market is fine either. UXPA’s 2024 survey found 37% of organisations had design layoffs, and the net staff reduction is the worst on their record, worse than 2009.

But look at what’s actually being cut. UK job ads have advertised salaries for UI designers down 20% year on year, and product designers down 7%. Same market, two very different jobs. The one being repriced is the one AI genuinely does now, which is making the screens.

Companies aren’t cutting the design job as it is. They’re cutting it as it was ten years ago, and honestly, they’re right to. That job is the one the tools do. What’s left is a different job.

What a model can actually reach for

Here’s the thing about a model. It’s trained on what already exists, and weighted towards what got shared a lot. That is a very good description of fashion.

So if you want a slightly better checkout, use AI. Really, use it. That work is a given now, and it isn’t what I’d want to be doing with your money.

But if you want a checkout nobody has seen before, the kind everyone copies for the next five years, that’s a different job. That’s designing for the need whether the technology exists yet or not. A model trained on what already exists can’t reach for that, because there’s nothing there to reach for.

Emily Brooke is the example I keep coming back to. Nobody briefed her. She was at a junction on her bike, wishing she could be five yards ahead of herself to warn the traffic. The laser light that came out of that ended up on Santander bikes across London. No brief, no ticket, no prompt. Somebody noticed a thing that didn’t exist and went and made it.

You cannot prompt your way to that, because the prompt is the answer. Somebody has to have the problem first.

The bit people misread

None of this is an argument against using the tools. I delegate to AI at almost every step of my job. I think of myself as the CEO of my own department: it does an enormous amount of the work, and none of the deciding.

Getting to okay used to take me a day, sometimes three. Now it takes an hour. That hour is real and I’m not going to pretend otherwise. But what you’ve got at the end of it is okay, and okay is where the job starts, not where it finishes.

The rest of the day is where I find out that the second screen assumes something the first screen never told anybody. That the thing your best customers do most often takes four taps. That the word on your button means something different to the people paying you.

None of that is visible at the end of the hour. That’s the whole problem with the hour. It looks finished.

Try this

Ask your team what they’d build if the technology already existed. Not what’s on the roadmap, not what a competitor shipped. What they’d build if the constraint were gone.

If the answers all sound like things that already exist somewhere, you’ve got a team that’s very good at executing. Which is fine, and you can probably get most of that from the tools now.

If somebody describes something that isn’t out there yet and can tell you who it’s for and why they’d want it, that’s the person worth paying for. That’s the painter.

The brush doesn’t paint. Somebody’s holding it.