A year ago, if a CV said the candidate was comfortable with AI tools, I noticed. It suggested someone curious, someone keeping up. That has worn off completely. The line sits on nearly every CV that crosses my desk now, and a claim everyone makes tells you nothing about anyone.
Most of what people mean by being good with AI has become fairly routine. Writing a usable prompt is ordinary now. So is cleaning up a rough draft, or getting a model to boil a forty-page report down to its point. Doing it well is about as impressive now as being handy with Excel. Somewhere in the last year, AI stopped being the thing that made a candidate stand out. It turned into something closer to knowing how to type.
So the real question in a hiring room has changed. Nobody senior asks whether a candidate can run the tool anymore. They assume it. What they want to know is what the person does with what the tool hands back.
That is where it gets human, and stays human. A model will give you an answer with total confidence, right or wrong, every time. The people worth hiring are the ones who can read that answer and feel the thing that is off in it before they can even say why. A lot of work looks finished. Sensing when it actually is takes something the software does not have, and it usually comes from having been burned a few times.
There is a harder version of this that matters even more. Most of the value shows up before anyone types a prompt, in deciding what the right question even is. AI will answer almost anything you put to it. It has no idea what is worth asking. Give me a candidate who can sit with a vague, badly framed problem, work out what actually needs solving, and only then open the laptop, and I will take that person over ten who can turn out a slick answer to the wrong brief. That instinct sorts people faster than any degree on the CV.
Step back, and the reason is simple. AI has made raw output cheap. Work that used to take a sharp junior half a day, and showed you how they thought while they did it, now takes twenty minutes and a prompt. That is a good thing. It also breaks the old shortcuts we used to judge people by. Speed does not mean what it did. Neat, finished-looking work does not either. What is left, and what has turned rare, is judgment. The willingness to make a decision, put your name against it, and carry the fallout, even when a machine did half the work. That is worth paying for.
If your hiring does not catch up to this, it fails in two ways, and I see both already. The first is obvious. You screen for AI like a keyword, and you end up with a team full of people who produce fast and rarely stop to ask whether the thing is right. You find out what that cost you when a real decision goes bad.
The second is harder to catch and does more harm. Once AI does your first cut, think about what it learned from. It learned from your old hires. That is the only pattern it has, so that is the pattern it looks for. A strong candidate who switched industries, or came up with a way nobody on your team recognises, gets dropped early. No person ever sees the name. Some of the sharpest judgment I have seen belonged to exactly those people, and an automated screen is built to miss them.
At Scrabble we use AI to bring us more names than we would have found on our own. We let it do nothing past that. The call stays with a person, and for us that is the whole point.
For an HR team, fixing this is smaller than it sounds. Change what you ask. Drop the question about whether they use AI. Everyone says yes, and you learn nothing. Ask about a time the tool got it wrong, and they caught it. Or a time they read what it gave them and chose to ignore it. Then listen to how they got there. Or hand them a problem that is deliberately vague and see what they reach for first. The ones you want tend to sit back and ask what is really being solved before they touch a keyboard.
For the record, I am not disowning AI or its usage. My point is narrower. The tool lifted the floor on what people can produce. It did not touch the ceiling on judgment. Companies that keep hiring for the floor will keep being caught out by those who turn out weak. The ones who learn to hire for the ceiling will build better teams while everyone else is still counting who can drive the software.
