If the junior tasks are gone, where does the senior judgement come from?

The numbers this year tell a strange story when you put them side by side. Stanford’s latest index puts developer employment for 22 to 25 year olds down nearly 20 percent since 2024, while their older colleagues’ headcount grew. At the same time, every account of what senior engineers now do lands on the same three words. Specification, review, orchestration. The human job is to oversee the agents and check the work.

Hold those two facts together and something does not add up.

The tasks we are automating first are the boilerplate, the simple fixes, the routine wiring. Those were never just cheap output. They were the training ground. They were how a junior spent five years quietly learning what good looks like, what breaks, where the bodies are buried. Judgement was the by-product of doing the unglamorous work.

So we are removing the apprenticeship for the one skill we have decided matters most. We tell new engineers their value is oversight, then hand them an environment that never builds the instinct oversight depends on. IEEE put it bluntly. Grads now have to slot in at a higher level almost from day one. From day one, with none of the years that used to earn it.

Some are already wrestling with it. IBM has reportedly rewritten its junior roles to drop the tasks AI now does and push early-career engineers toward problems and people instead. That is one answer. It is not the only one, and most teams have not even asked the question.

This is not a reason to stop using the tools. The productivity is real and it is not going back. But it does change what a team has to be deliberate about.

If review is now the job, then teaching people to review has to become an explicit act, not something we hope rubs off. That means junior engineers reading agent output next to someone who can say why this is wrong and that is fine. It means making the reasoning visible, not just the result. It means treating judgement as something a team builds on purpose, because the old path that built it for free has been paved over.

The uncomfortable version of the question is this. A checker who cannot tell a good answer from a plausible one is not a check. They are a rubber stamp with a job title. We are on course to produce a lot of them unless we are honest about where judgement actually comes from.

What is your team doing to build the instinct, now that the old training ground is gone?

The follow up data suggests this is not mean reverting, as Fortune’s June coverage of the Stanford and ADP research set out.