Stanford tracks a 4.4% employment drop for young workers in AI-exposed jobs

Stanford HAI’s October 21 account of its labor-market tracking highlights a warning for workers entering AI-exposed careers: employment among 22- to 25-year-olds in software development and customer service has fallen, with the decline recently reaching 4.4%. The finding comes from payroll records, but it does not establish that AI caused the losses.

The figure describes an existing trend, not a newly measured October decline. France 24 reported on October 6 that US employment among workers ages 22–25 in the most AI-exposed roles was down 4.4% year over year in August. Its coverage of the Canaries Dashboard findings said employment in those roles had contracted every month since October 2023.

Payroll records provide an early warning

The Canaries Dashboard is a project of the Stanford Digital Economy Lab and ADP Research. It draws on anonymized payroll records from ADP Research’s parent company and groups workers by occupations with high or low AI exposure. That design lets researchers examine specific parts of the workforce rather than rely only on economy-wide totals.

Stanford describes the dashboard as an early-warning tool. Bharat Chandar, the lab’s research director for AI economics and policy, says the researchers examine narrow sections of the labor market to detect emerging changes in people’s work. The dashboard builds on the research paper Canaries in the Coal Mine?

−4.4%Employment change, ages 22–25

US employment in the most AI-exposed roles fell 4.4% year over year in August, according to the Canaries Dashboard findings covered by France 24.

A gender signal faces a separate test

The dashboard also shows slower job growth for young women than for young men. Chandar cautions that ADP Research’s data may not be representative enough to confirm that broader pattern. The concern is whether the workers captured in those records adequately reflect the population needed to draw a wider conclusion.

That preliminary result aligns with other research indicating that women disproportionately work in AI-exposed fields, Stanford says. But exposure and measured employment outcomes are different questions. Chandar notes that longer-term government statistics show no evidence of a gender gap, leaving the dashboard’s apparent divergence as something to monitor rather than a settled finding.