© Nink AI Engineering GmbH

The Junior You Didn't Hire

For eighteen months the consensus was that fewer people meant faster growth. US public companies shrank their white-collar workforces on the theory that agents would absorb the difference. Last month the Wall Street Journal reported the reversal: CSX, Alphabet, Snap-on and Booz Allen Hamilton all telling investors they intend to add people again.

The interesting part is not the numbers. It is the stated reason.

Sarah Franklin, CEO of the HR platform Lattice, describes companies that stopped hiring entry-level staff believing AI agents would pick up the slack, and have since realised humans are needed to work alongside them. Her line is the one worth keeping: "Just because you have coding agents doesn't mean you're not hiring engineers."

I recognise the mistake, because I have been running the experiment on myself. I have ten agent systems doing work I used to do by hand — tax collection, research briefs, review preparation, documentation. Not one of them removed a role. What they changed is what the role consists of.

The tax pipeline files nothing; it produces a pack a human signs. The review system prepares a review; I still deliver it. The option study stops at the trade-off and hands it back. Every one of them ends at a judgement I did not delegate, and that is by design rather than by limitation.

Which is the part a hiring freeze gets backwards. The output of an agent fleet is not fewer decisions. It is better-prepared ones — and someone still has to be there to make them.

The honest counterweight comes from Paul Osterman at MIT, quoted in the same piece: "Do we need more people? Do we need less people? We have no idea. No one has any idea."

He is right. Anyone selling you certainty about this, in either direction, is selling.