I took a look at some of the stats from our projects for June and found a gem. In that month, we delivered a complete, branded design system in a single day. A full product dashboard with mobile views in 26 hours. An interactive onboarding prototype in three.
Those aren’t estimates that would’ve been possible a year ago. And while it sounds impressive, the takeaway is actually another question.
What does implementing AI actually do to a design studio?
Honestly: it depends on which studio you're asking.
The middle is collapsing.
AI is very good at the repeatable parts of design work. Wireframes. User flows. Component generation. Copy variants. The kind of output that used to take a junior designer two days now takes an afternoon, sometimes less.
For studios that compete on execution speed, that's a real threat. If your value proposition is "we make things quickly and cheaply," you're in a race where the finish line keeps moving and your competitor never sleeps.
And yet, something else is happening at the same time. The ceiling on what's possible is rising.
When the commodity work compresses, senior judgment becomes a scarcer resource. The ability to generate a button component or a responsive grid still matters, but now, you can outsource it. Where you really add value is knowing what the product actually needs, why, and what it’ll cost the business if you get it wrong. You need context, and nuance.
Contrary to what some are saying, I don’t believe that's something you can prompt your way to.
What AI can't replicate.
It took four years to design the product infrastructure for CloudKitchens. One unified design system powering dozens of applications, serving chefs, couriers, regional managers, and analysts - all with different devices, environments, and pressure levels.
A chef mid-service needs a tap target you can hit with a knuckle. A regional manager pulling margin data across fifty locations needs refined controls and information hierarchy that doesn't add cognitive load to an already complex job. The same design system had to feel native to both.
No AI tool designed that. No AI tool could have. Not because it lacks the capability to generate decent screens (it doesn't), but because the judgment that shaped those decisions came from years of working inside complex products, understanding what breaks under operational pressure, and knowing which design choices compound into problems six months later.
That's what I mean by the gap getting wider.
A studio using AI to generate a component library and ship it as a design system is faster than it was before. They’re not doing the same work we are.
Speed is only valuable when it's pointed at the right problem
Here's what the output numbers from last month actually represent. We delivered a design system in a day because we'd spent months building Wideset - our own component foundation - before any of those client projects started. Roughly 80-90% of what we shipped in May was reusable infrastructure we'd already built and refined.
We built the foundation, AI helped us execute against it faster - but it didn’t create it on its own.
That distinction matters. Speed without a strong design foundation is just faster mediocrity. The frameworks, the judgment calls about component architecture, the decisions about what to systematise and what to leave flexible… those came from experience, not from a model.
What AI gave us was the ability to compress the distance between a design decision and a working, testable implementation. A prototype that would have taken a week now takes a day. A client can react to something real instead of something theoretical.
What does that change? The quality and ease of feedback for one. It changes the level of conversation with the client. It changes just how much you can learn in a sprint.
But only if you started with the right question.
The studios that will struggle
I'm not worried about AI replacing senior designers. I'm watching something more interesting, and more uncomfortable, for a lot of agencies.
The studios that will struggle are the ones that were already operating below a certain threshold. If the majority of your work was execution, rather than judgment and strategy, and you’re counting your value in hours rather than what happens during them - you’re in trouble.
For those studios, AI doesn't level the playing field. It removes their position on it.
The studios that will do well are the ones that were already doing things AI can't replicate: deep discovery work, researched brand strategy, systems thinking, understanding the business context behind a design decision, knowing when to push back on a brief.
For those studios, AI is the best thing that's happened in a decade, because it removes the friction between thinking and making. It lets senior designers spend more time on the problems worth solving and less time on the work that was always below their actual capability.
At least, that's what we've found.
What this means if you're buying design work
If you're evaluating a studio right now, ask them what they've built that doesn't belong to any one client. Ask what their reuse rate looks like across projects. Ask what they'd do differently on a brief than an AI tool running the same prompt.
If they struggle to answer any of those, think again about what you’re really buying. Is it judgment, or execution? And remember - execution is getting cheaper every month.
The gap between average and great in design has always existed. AI didn't create it - but it did make it a lot harder to hide.
Reach out to us if you're looking for a studio that builds on strong foundations for enterprises and scaleups.
.png)
