Revolut's Designers Got 3x Faster. Faster at What?
Revolut's CEO says his designers got 3x faster by letting an LLM build with their design system. Designer Rasmus Andersson says "designing fast" isn't even a thing. I think the real question is what exactly got faster.
The AI built the screens faster. That's real, and it's useful. But in my experience, building the screens was never the slow part of design.
The slow part is deciding
In enterprise products, the time goes into decisions. What happens when the data is missing? Who is allowed to undo this action? What does a screen-reader user hear when the form fails? Answering those questions takes research, conversations and sometimes a few disagreements.
Once those answers exist, the screens come quickly, with or without AI. So when a team says it designs 3x faster, I want to know: did the decisions get faster, or only the screens?
Fast screens without decisions multiply mistakes
If you speed up building but skip deciding, you don't save time. You move the problem later, and you make more of it.
That happens because the AI builds from your design system. Every rule in the system gets repeated in every screen it produces, and so does every missing rule. If the system has no pattern for errors, empty states or accessibility, the AI will invent one, slightly differently each time, across a hundred screens. Then someone has to find and fix all of them.
So the design system is no longer just a library. Once AI is building with it, it's doing the designing.
How to make speed actually help
Speed is valuable when it comes after the decisions, not instead of them. Here's what I would do:
- Decide first, then build fast. Measure how quickly the team reaches an answer it can trust, not how many screens it produces.
- Put the decisions in the system. Write the states, content rules and accessibility requirements next to the components, so the AI follows them instead of guessing.
- Use the speed to test earlier. A working prototype can be tested with a keyboard, a screen reader and zoom on day one. That's where faster building pays off: you find problems sooner, while they're cheap to fix.
What "fast" should mean
Fast should mean getting to the right answer sooner. AI has made building cheap, which makes deciding well the most valuable part of a designer's job. That's also what designers should show when leadership asks about speed: not how many screens we made, but which decisions made all of them right.