Daniel Birch

Shape, texture and composition.

Speeding up design crits. I’ve been wondering lately how useful AI actually is when it comes to helping designers design.

Not the production side - we already know it’s good at that. It can generate screens, tidy things up, get you to something polished much faster than before. But the thinking side is more interesting to me. Can it help with purpose, layout or enerating something quickly enough to spark a better conversation?

That’s what I wanted to explore.

Design crits were getting a bit… predictable. We run weekly design crits in the team. They’re a good space. People share work, challenge each other, build confidence. When they’re working well, you get proper discussion—different perspectives, strong opinions, all of that. But recently, they’d started to feel a bit samey. You could almost predict how they’d go:

  • Someone points out a usability issue

  • Someone references accessibility

  • Someone brings up a known heuristic

All valid. All important. But it started to feel like we were just going through the motions. Like we were ticking off best practice rather than really thinking.

So I wanted to shake it up a bit.

What if AI did the first pass? The idea was simple. Instead of starting from scratch in the crit, what if we let AI do the initial critique - and even propose a solution?

So I took Spotify’s new user sign-up page and gave it a prompt along the lines of:

“Critique this page based on how well it drives sign-up, then redesign it based on your recommendations.” Then I did the same with Deezer, which (until recently) had a pretty similar setup. What came back was… solid.

  • Clear structure

  • Logical hierarchy

  • Strong CTAs

  • Sensible use of brand

  • A neat list of benefits

If you put that in front of a product manager or a stakeholder, they’d probably be pretty happy. It all made sense. It was clean, purposeful, and well-reasoned.

But also… a bit boring. That was the thing. Both outputs felt safe. They followed patterns you’ve seen a hundred times before. They looked like they should work. They ticked all the right boxes. But they didn’t really say anything. No edge. No personality. No sense that this was this product versus any other. And that’s where it got interesting. Because AI is doing exactly what it’s supposed to do - it’s pulling together best practice. It’s giving you a well-optimised version of what already exists. But that’s not the whole job of design.

The crit got way more interesting. Bringing this into the crit changed things almost immediately, because the basics were already covered. We didn’t need to spend time talking about:

  • Whether the CTA was clear

  • Whether the layout made sense

  • Whether it followed known patterns

AI had already done that. So the conversation moved on. It became more about:

  • “Is this actually interesting?”

  • “Does this feel like the brand?”

  • “What would make someone choose this?”

People had stronger opinions. The discussion had more energy. It felt less like a checklist and more like actual design thinking. At times it was a bit heated - but in a good way.

It highlighted something important. What this really showed is that AI is very good at getting you to “good”. Clean. Usable. Logical. All the things we’ve spent years defining as best practice. But “good” isn’t the same as memorable. Or distinctive. Or interesting.

And that’s where designers come in.

It also made me reflect on how we’re working. There’s a risk here. If we lean too heavily on:

  • Design systems

  • Established patterns

  • AI-generated solutions

We end up levelling everything out. Everything works. Everything is consistent. But everything starts to feel the same. Following guidelines is important - it gets you to a solid baseline. But it’s not where the interesting work happens.

A better way to use AI in crits. What worked about this approach was the shift in focus. Let AI handle the baseline.

Let it:

  • Point out the obvious issues

  • Apply best practice

  • Generate a “safe” version

Then use the crit to go beyond that.

To ask:

  • What’s missing?

  • What could this be?

  • How do we push it further?

It turns the crit from a review session into something more exploratory.

And honestly, it’s just more fun.

Final thought. AI is brilliant at giving you a starting point. But it naturally pulls you toward what already exists. As designers, we need to decide if we’re happy stopping there. Because “good” is easier than ever to achieve.

But the interesting stuff -the bits that stand out, that people remember, that actually feel like a brand - that still takes people in a room, challenging each other and pushing ideas further.

And for me, that’s still the best part of the job.

This is a footer, but by now you are bored, so nothing more…

© danielbirch 2026