This commentary originally appeared in K. Sudhir’s Substack newsletter The Slow Part. The views expressed are the author’s own.
Many senior people are noticing the same thing lately: their junior staff are doing better work than ever—cleaner drafts, faster models, fewer mistakes—because AI now handles the grinding first pass, the rough draft that used to eat a junior’s week. And then, often in the same breath, comes the hesitation: they’re not quite sure how these young people are going to actually get good. Not good with the AI tools. Good the way the senior person is good, in the way that took them 15 years.
That hesitation, I think, is pointing at something real. In a recent BCG study, more than half of senior executives named the slower development of junior talent as a concern. It’s easy to hear it as the story everyone’s already discussing—AI thinning entry-level hiring, the bottom of the ladder narrowing. That’s true, and well documented. But the hesitation is about something quieter, sitting underneath the jobs question: what the junior work was doing besides producing output.
Think about what the junior work actually was. The analyst’s rough financial model wasn’t a slower version of the spreadsheet the machine now produces in seconds. It was how she found out which assumptions actually move a business—by building it wrong, seeing where it broke, and building it again. The associate’s clumsy first markup of a contract wasn’t just a slower way to flag the problems. It was how he learned which risks are real and which ones only look dangerous. The work made two things at once. It made the deliverable. And, less visibly, it made the person.
Practice may also feel slow and repetitive, but it quietly forces you to make the calls that judgment is made of: which assumption changes the answer, which risk is real, which draft is merely fluent and which one is actually right.
That second product—judgment, the feel for the work—is the one that matters most over a career, and it’s the one nobody could ever buy directly. You got it by doing the work, including the parts that were slow and repetitive and beneath you. Which is why AI raises a question we haven’t really had to ask before. For the first time, the output can be produced without the person producing it. The deliverable arrives; the practice that used to come with it doesn’t.
And here’s the part I find genuinely interesting, because it explains why so few organizations are reacting to any of this. The practice that formed people was never on anyone’s invoice. It came free, riding along inside work the firm was already paying for. Nobody ever had to price it, budget for it, or defend it, because you couldn’t get the output without it—the two were welded together. AI is the first thing that pries them apart. And the moment formation stops being a byproduct you get for free and becomes something you’d have to provide on purpose, it runs into a hard fact: it was never a line item, so nothing in the system notices when it goes missing. There’s no cost that spikes, no metric that dips this quarter. The productivity shows up immediately and visibly. The missing formation shows up years later, quietly, as people who were supposed to have become senior and somehow didn’t.
None of this is an argument for romanticizing drudgery. “It builds character” has always been the favorite excuse for keeping people in toil, and plenty of junior work taught nobody anything—re-keying numbers from one system into another built no judgment, and the machine should take it. The difference that matters is between toil and practice. Toil just consumes the hours. Practice may also feel slow and repetitive, but it quietly forces you to make the calls that judgment is made of: which assumption changes the answer, which risk is real, which draft is merely fluent and which one is actually right. The danger isn’t that AI takes the toil. It’s that, in taking the toil, it also takes the practice—and nothing gets put in its place.
So the question I’d put to anyone rolling out AI, right next to “what can this take off our hands?”, is one that’s easy to skip: what did people learn by doing this, and where will they learn it now?
And it isn’t only the senior people who might feel this. I hear the other side of it from my own students. If you’re early in your career, you may already sense it: you’re turning out work you couldn’t quite have produced alone, and no one seems to expect you to. The tool makes you look capable now. The quieter question, the one that’s harder to say out loud, is whether you’re actually becoming capable.
There are ways organizations can protect how people develop judgment even as they lean on AI—but all of them start with recognizing that this is happening at all. That’s the first move: to stop treating it as a side effect of automation and start treating it as something to design for. Because organizations are very good at counting the output AI produces, and much worse at noticing the practice it quietly removes.
Subscribe to K. Sudhir’s The Slow Part.
“The Yale School of Management is the graduate business school of Yale University, a private research university in New Haven, Connecticut.”
Please visit the firm link to site
