AI adopters shifted toward senior workers. That is not a layoff wave.
A new Stanford working paper follows foreign affiliates of multinational companies in 41 countries and identifies generative-AI adoption from company job advertisements that involve GenAI. It adds a more direct adoption signal to a map dominated by exposure scores.
By March 2026, the paper’s matched, instrumented comparison estimates a 1.9-percentage-point lower junior share at adopting affiliates than at matched controls. Senior employment was 6.7% higher. The estimated 2.5% fall in junior employment was not statistically precise, and total employment was estimated 3.3% higher at marginal significance.
What changed
This is not a count of layoffs. The main result is a workforce-composition shift: firms in the study appear to grow senior roles faster while junior representation falls. That makes an entry-route warning more concrete than an exposure map alone, but it does not show that AI has cut junior jobs across the economy.
Why the claim remains bounded
The authors use a company GenAI job ad as a proxy for adoption, then instrument adoption with peer adoption around the parent company’s headquarters. The design is stronger than a simple before-and-after comparison, but it depends on that peer effect reaching the affiliate only through the parent’s adoption. The data also overrepresent managers and professionals, and the main analysis covers larger foreign multinational affiliates.
How this fits the broad U.S. check
Yale Budget Lab still finds no clear aggregate U.S. AI labor-market footprint through August. The two results can coexist: one tracks a selected set of adopting firms and their workforce mix; the other tracks broad U.S. indicators. The evidence now supports a live concern about junior representation inside AI-adopting multinationals, not an economy-wide layoff headline.