The Signal · Column 068

The Hiring Slowdown Arrived Before The Adoption Did

Andrei van Noordt · 25 September 2026 · 2 min read

Goldman Sachs went looking for how far artificial intelligence has actually travelled inside working companies and reported that only four countries have crossed twenty percent adoption, even as technology hiring slows. Two findings in one sentence, pointing in opposite directions. The tools are still early in their spread. The hiring plans are being written as though the spread is already finished.

The part most people skip is the order of events. A slowdown in technical hiring is normally a lagging signal. It follows the productivity, arriving after the machines have visibly taken over some of the work. What this data describes is a slowdown arriving first, ahead of the adoption that is meant to justify it. Very few companies are cutting because artificial intelligence replaced the work. They are cutting because they expect it will, and an expectation is far cheaper to act on than a deployment.

Most hiring freezes this year are funded not by artificial intelligence but by the belief in it.

This is how a labour market misprices people. In the large majority of markets sitting below that adoption line, the work still has to be done by humans, and the humans who could carry a company across the line are caught in the same freeze as everyone else. The teams that would build the adoption are being thinned by the anticipation of it. Meanwhile the small number of companies genuinely past the threshold are competing for the engineers who got them there, out of a pool that the freeze itself has made shallower.

So the repricing is not a general one. Generic technical headcount is getting cheaper on a promise. The specific people who can take a business from pilot to production, the ones who have done an integration end to end and know where it breaks, are getting more expensive on evidence. Those two curves are moving apart right now, and a hiring plan written on the first one will be paying the second one's prices by the time it needs those people.

For hiring leaders, the practical read is simple. Separate your headcount plan into what deployed artificial intelligence has actually absorbed and what you merely expect it to absorb, and only cut against the first column. The second column is a forecast, and forecasts do not ship anything while your competitors are hiring the people who could.