A few years ago, the bottleneck in most work was production. Writing the memo, building the model, drafting the code, making the deck — the doing took time, and the people who could do it quickly and cleanly were valuable precisely because that competence was scarce. That world is ending. When a capable person with good tools can produce a competent draft of almost anything in minutes, competence stops being the thing that separates candidates. Everyone's output starts to look the same.
So what's left? The part that never automated: deciding what to build, when it's actually right, and when to throw it out and start over. I've come to think of that as judgment, and it's quietly becoming the most important thing I screen for when I hire.
Output is cheap; knowing which output matters is not
AI is extraordinary at generating options. It is far less reliable at telling you which option is correct for your specific situation, your constraints, your customers. That gap is judgment. It's the ability to look at ten plausible answers and know which one is right — and, more importantly, to know when all ten are wrong and the real move is to reframe the question.
You can't fake that in an interview with rehearsed answers, and a machine can't produce it on your behalf, because it depends on context the machine doesn't have: what your company is actually trying to do, what you've tried before, what will quietly blow up in six months. The people who have judgment are the people who've been close enough to real decisions, and their consequences, to develop an instinct for them.
How I try to see it
Résumés are almost useless here, because judgment doesn't show up in a list of titles. What shows it is work — real work, aimed at a real problem, where the person had to make choices and can explain why. Show me something you built and I'll learn more from ten minutes of "why did you do it this way and not that way" than from an hour of behavioral questions. The explanation is the signal. Good judgment sounds like someone who considered the alternatives and can tell you, without defensiveness, why they ruled them out.
This is the whole premise behind what I'm building at Provieo. When anyone can generate a flawless-looking application, the scarce and honest signal is a body of demonstrable work plus the reasoning behind it. That combination is very hard to manufacture and very easy to recognize once you know to look for it.
What this means if you're early in your career
The advice I'd give someone starting out today is almost the opposite of what I'd have said a decade ago. Don't compete on how fast you can produce — the tools have made that a losing race. Compete on the quality of your decisions. Use AI aggressively for the rote work, then own every judgment call on top of it: what to make, what to cut, what "good" means here. Build things, and get in the habit of explaining your reasoning out loud. That muscle — deciding well and defending it honestly — is the one that compounds, and it's the one no model is going to take from you.
Machines will keep getting better at answering the questions we give them. The enduring human advantage is knowing which questions are worth asking. That's what I'm hiring for, and I suspect it's what most thoughtful managers will be hiring for too.