Banking Is an Apprenticeship Business — and We're Burning the Ladder

I am a product of the work we are now automating away.

Early in my career I built collections models for a global finance operation — bringing down delinquent receivables, high-volume, low-prestige analytical work at the bottom of the institution. Exactly the category AI absorbs first. The models improved efficiency, and that was never the real output. The real output was what the work did to me. Ten thousand repetitions of looking at receivables taught me what normal looks like, where the data lies, which exceptions matter and which are noise. Nobody can hand you that in a deck. It accretes through volume, through being wrong enough times to develop an instinct for right. Later I built analytics centers across four countries and recruited hundreds of analysts, and I watched the same conversion happen at scale. Routine work in, judgment out.

The Contradiction Nobody Has Noticed

Now look at 2026. Banks are cutting junior analyst classes by as much as two-thirds. And they are sourcing the majority of their AI talent from those same junior cohorts. Shrinking the entry pool and drawing their most important new capability from it, at the same time. The contradiction hasn't registered because both decisions are locally rational. Cutting juniors saves money this year. Hiring AI talent from juniors solves a problem this year. The collision is a decade out, which in a quarterly business is the same as invisible.

The logic of the cutters runs: AI does the grunt work now, so we need fewer people doing grunt work. True, and it misses the point of the grunt work. The grunt work was never mainly about getting the grunt work done. It was the manufacturing process for senior judgment. Today's managing directors are the people who did, 20 years ago, exactly the work AI now does. They did not develop judgment in a leadership program. They developed it in the volume — the thousands of ordinary cases that taught them to recognize the extraordinary one. Remove the volume and you have not cut headcount. You have unplugged the machine that produces your future leadership. The question almost nobody has asked is simple. Where do the managing directors of 2036 come from?

The Apprenticeship Was an Accident

The obvious response — keep hiring juniors, let them do the grunt work — is also wrong. The honest truth is that the apprenticeship was never designed. It was a byproduct. Juniors built judgment by accident, as a side effect of high-volume work that needed doing anyway. It was an absurdly expensive way to manufacture senior people. We never had to admit that, because we never had an alternative. AI has now removed the work the apprenticeship was riding on. You cannot keep the old apprenticeship. The thing it was attached to is gone.

Which means the institutions that survive this will have to design, deliberately, the thing they used to get for free. That is hard. It is also a real opening, because almost nobody is doing it. If judgment used to come from ten thousand routine cases, and the routine cases are now handled by machines, then development has to move to where judgment-building work still lives. There is plenty of it. Put juniors in the exception queue instead of the routine queue — the hard, ambiguous, model-defeating cases. Put them in the review loop on the AI's own decisions, where the interesting failures surface; learning to spot a confident-but-wrong model output is exactly the judgment the next decade requires. Rotate them through the messy cross-functional work no model touches. The reps still exist. They have moved, and they have to be built on purpose now instead of harvested by accident.

Who Trains the Class of 2036

The institutions that treat the apprenticeship as something to rebuild will own the next generation of leadership in their industry. They will have senior people in 2036 because they engineered the pipeline that produces them. The institutions that stop hiring juniors, bank the savings, and move on will discover something around 2032 that does not appear in any business case being written today: they outsourced their succession plan to their competitors. The judgment they stopped manufacturing will have to be bought at a premium from the few who kept building it — if those people are even available at all.

Senior judgment cannot be hired in bulk when you suddenly need it. The people who have it are the ones who did the work that builds it, and that work is being eliminated across the industry right now. The savings are immediate and visible. The bill arrives in a decade, as an absence — the managing directors who were never made. Banking has always been an apprenticeship business. The question is whether anyone rebuilds the apprenticeship now that the old one has been automated out from under it, or whether we burn the ladder for a quarterly saving and act surprised, ten years on, that no one can climb.