Is AI Closing the First Rung of the Career Ladder?

1 The claim and its sample

A working paper revised August 12, 2026 by Erik Brynjolfsson, Bharat Chandar and Ruyu Chen of the Stanford Digital Economy Lab documents six facts about the labor market after the widespread adoption of generative AI. The data is high-frequency administrative payroll from ADP, covering millions of U.S. workers through June 2026.

That is a strong sample for one purpose and limited for another. Payroll records show who was paid by which employer and when, not why a manager chose one candidate over another.

2 A relative change, not a headcount

The headline number is a comparison. The authors report employment of young workers aged 22 to 25 in AI-exposed occupations now stands 19% below where it would be had it kept pace with less-exposed peers, while experienced workers show no comparable gap.

Nineteen percent is the size of a gap between two groups of workers, not a count of jobs removed from the economy.

3 Hiring, not firing

The paper reports the divergence operates primarily through reduced hiring of young workers rather than increased separations, and that adjustment is occurring through employment rather than base compensation. Declines concentrate where AI usage substitutes for human tasks; where usage complements workers, employment is flat or rising.

4 Alternatives the authors tested

The authors state the divergence is not explained by several prominent alternatives. It persists when excluding technology firms and computer occupations, when controlling for interest-rate exposure and remote work, and across alternative measures of AI exposure. The lab also published a February 2026 note on interest rates and timing.

5 Where the pattern weakens

The paper is candid about attenuation. The patterns attenuate when controlling for education, some divergent trends predate generative AI, and the divergence is more pronounced in the ADP sample than in national survey benchmarks, with only some evidence of consistent patterns in government administrative data.

6 What the data does not show

The authors state plainly that they find no evidence of widespread, economy-wide job displacement, and interpret their results as early, descriptive indicators - canaries in the coal mine - rather than causal estimates. Nothing establishes that AI caused the gap rather than accompanied it.

Three limits follow. The finding concerns young workers inside exposed occupations, not total employment. Payroll data cannot show where unhired young workers went, since a person who never appears is not tracked. And no employer is shown choosing an agent over a candidate; the mechanism is inferred from where the gap concentrates.

7 Bottom line

The reported figure is a 19% relative gap for workers aged 22 to 25 in AI-exposed occupations, not a national job-loss count.

The described channel is reduced hiring rather than separations, and the paper reports no evidence of economy-wide displacement.

The authors call the findings early descriptive indicators rather than causal estimates, and the pattern attenuates on education controls and in broader benchmarks.