How to Interview a Designer Now

The signals hiring managers relied on for a decade stopped carrying information. Polished case study, working AI-powered prototyping in a weekend , tidy design system, a portfolio that looks like it came out of a studio. All of that is now available to anyone with a weekend and a subscription.

I am not going to argue about whether that is good. It happened. The question is what you ask instead, and most interview loops have not caught up.

Stop Grading the Artifact

A portfolio used to be evidence of two things at once: this person can produce work at this level, and this person cared enough to finish. Both inferences were reliable for a long time and neither one survived polish becoming free. A demo has the same problem. A working prototype used to prove somebody had the skill to build it, and now it proves they had a weekend.

None of that makes the portfolio useless. It makes it a starting point for questions rather than an answer to them. Look at the work, then spend the whole conversation on what is not in it.

Ask What They Deleted

This is my first question and it separates people faster than anything else I have tried. What did you remove from this, and who was unhappy about it?

Adding is easy to justify and easy to generate. Deletion requires an opinion about what the product is for and the willingness to defend it against someone who wanted the thing you cut. Nobody deletes by accident, so every deletion has a reason attached and the reason is where the judgment lives.

The answer I want sounds like: we cut the tutorial because the product is opened from a link and the first ten seconds are all the attention we get. The answer that tells me to keep digging is a list of everything they built.

Ask Where the Number Came From

Every portfolio now has percentages in it. Most of them cannot survive one follow-up question, which makes the follow-up question extremely efficient.

I can say that audio failed in about forty percent of Ring-Rival sessions and is under two percent now. If you ask how I know, there is an answer: browsers block audio until the user interacts, the game was starting its audio context on load, and moving that behind the first tap fixed it. Nothing errored the entire time, which is why nobody would have reported it.

Ask how it was measured, what it was before, and how they found out it was a problem at all. A real number has a story about detection attached to it. An invented one has a story about success.

Ask About States, Not Screens

Screens are what got cheap. States are where products actually break.

On HerbaLink, practitioner verification looks like a checkmark and an upload form until you build it, at which point it is submitted, under review, approved, rejected, expired, and resubmitted, plus the question of what happens to an existing booking when the credential behind it lapses. That question has a right answer and it is not a visual one.

So take any feature in their portfolio and ask for every state it can be in, and what the user sees in each. Designers who have shipped will enumerate them without hesitating. Designers who have only drawn will produce the happy path and then start improvising.

Give Them a Real Problem and Let Them Use AI

Banning AI in the exercise measures a skill the job no longer contains. Do the opposite. Hand over an actual unresolved problem from your backlog, give them thirty minutes, tell them to use whatever tools they want, and ask for no deliverable at all.

Then watch two things. What they ask before they start producing anything, because the questions reveal how they frame a problem and questions cannot be generated for you. And what they overrode, because the interesting part of working with a model is the moment you look at a competent, reasonable output and decide it is wrong for this specific product.

Somebody who accepts the first plausible answer will do that on your product too, every day, at speed.

The One Thing I Would Not Skip

Ask what they would fight for. Not what they believe about design in general. What in this specific piece of work they would defend if a senior person told them to change it, and what happened the last time that occurred.

Conviction is the trait that survived. Producing options is free, defensible-looking work is free, and the scarce thing is somebody who can tell you why the reasonable answer is wrong here without falling back on a rule they read somewhere. That has never been visible in a portfolio, and now it is the only thing worth interviewing for.