The headline is resilience; the hidden signal is entry

The paper does not find a general employment collapse across the economy. Its strongest pattern appears when the workforce is separated by age inside occupations with different levels of AI exposure. Workers ages 22 to 25 in the most exposed jobs fall behind while older workers in the same kinds of roles do not show a comparable decline.

The authors estimate a 19% gap relative to the path young exposed workers would have followed had they kept pace with less-exposed peers. That is a comparison of trajectories, not a claim that 19% of all young workers lost jobs.

Hiring is the pressure point

The divergence appears primarily in new hiring rather than separations. That makes the disruption less visible than a layoff wave but potentially more consequential over time. A profession can retain experienced staff while sharply reducing the number of beginners allowed to enter.

The study also distinguishes substitution from complementarity. Young-worker declines are concentrated where AI is more likely to substitute for labor, while employment in complementary roles is flat or rising, particularly for experienced workers.

Treat the result as an early warning, not a verdict

The analysis is descriptive. It does not prove that AI alone caused the employment gap, and the authors identify pre-existing trends, sample differences, and sensitivity to education controls. ADP's payroll population also does not perfectly represent the national workforce.

The appropriate response is better measurement. Employers and policymakers should track entry-level postings, hires, wages, training slots, and promotion pathways by age and occupation. Aggregate employment can remain calm while the next generation's route into expertise narrows.

  • Separate hiring from layoffs when measuring AI labor effects.
  • Track substitution-heavy and complementary uses of AI independently.
  • Preserve paid pathways that let newcomers acquire expert judgment.
  • Update the estimates as broader national data become available.
Primary trail

Go to the source

Read the evidence behind this analysis. External links open in a new tab.

Stanford Digital Economy Lab — Canaries in the Coal Mine Stanford Digital Economy Lab — Revised working paper