We Only Hire Designers Who Use Claude. No One Else Gets Through.
We changed our hiring bar. If a designer does not work with AI, and specifically does not work with Claude as part of how they think, we do not move forward. Portfolio, years of experience, and taste do not override it.
That sounds like a gimmick, and I understand why. So here is the actual reasoning.
The filter is not about the tool
I do not care that a candidate knows the keyboard shortcuts. Tools change. What I care about is what the answer reveals.
Designers who have genuinely absorbed AI into their process have, without exception, done two things: they went looking for it before anyone told them to, and they kept going after the first week, when it stopped being magic and started being work. That combination, curiosity plus follow-through, is the thing we were always trying to hire for. We just never had a clean way to test for it.
Now we do. It is the cheapest signal in the interview.
What "uses Claude" actually means to us
Not "generates mockups." That is the least interesting thing it does.
Research synthesis. Forty user interviews, read properly, in an afternoon. Not summarized into mush, interrogated. What contradicts our assumption? Which quotes argue against the direction we already like?
Options at zero marginal cost. Twelve directions instead of three, produced fast enough that being wrong is cheap and killing your own idea stops feeling expensive.
Prototypes that actually run. A designer who can describe an interaction to Claude and get a working thing in the browser has removed the single largest source of loss in our work, which is the gap between what the mockup implied and what the build turned out to be.
Real inputs, wired in. Pulling actual product screens and design context into the tool rather than working from memory and dribbble. We wrote about Mobbin's MCP server for exactly this reason.
The obvious objection
"You are going to hire prompt jockeys with no taste."
It is the opposite, and this is the part people get backwards.
When producing a competent screen was expensive, production skill was a real differentiator, and you could build a career on it. That era is over. When twelve options cost the same as one, the scarce skill is not making them. It is knowing which one is right and having the nerve to delete the other eleven.
AI does not replace judgment. It removes every excuse for not having any.
A designer without taste using Claude produces more mediocre work faster. That is visible in about four minutes of conversation. The filter does not let them through, it exposes them sooner.
How we test it
Not with a portfolio review. Portfolios show finished work, and finished work has been edited by time.
We run a live exercise on a real problem, with AI explicitly allowed and encouraged. Then we ask about the parts that are not on screen. What did you throw away? Where did the model give you something confident and wrong? What did you decide it was wrong about, and how did you know?
The answers separate people immediately. Strong candidates have a running argument with the tool. Weak ones accept the first plausible output and present it as a decision.
Why this matters if you are the client
You are not paying us for hours. You are paying for the quality of decisions and the speed of getting to them.
A team that works this way explores more of the possibility space before committing, tests against real users earlier, and hands you prototypes instead of pictures. The work gets better because the loop got shorter, not because anyone typed faster.
We would rather be honest about the bar than pretend it is business as usual. It isn't.
Contrast is a psychology-driven UX studio for SaaS companies and startups, run by a team that builds this way by default. Book a call and see what a shorter loop does to your product decisions.