Stanford–ADP update finds Gen Z entry jobs weakening where AI exposure is highest

Stanford Digital Economy Lab and ADP updated their Canaries dashboard to show entry‑level employment among 22‑ to 25‑year‑olds has weakened most sharply in occupations the researchers classify as highly exposed to generative AI. The dashboard also shows that women in that age group have experienced slower employment growth than men, a pattern the authors say is largely explained by occupational sorting even as AI exposure is a contributing factor.

The update, published July 22, 2026, overlays occupational measures of generative‑AI exposure on high‑frequency payroll data to produce real‑time indicators of hiring and employment at the “canary” level. According to the Stanford and ADP team, entry‑level employment in AI‑exposed roles has lagged relative to less exposed occupations, while the gender gap in early‑career employment narrows considerably once differences in which occupations young men and women enter are taken into account.

The finding sharpens a trend HR leaders have been monitoring all year: as firms adopt generative AI tools, early‑career hiring is shifting unevenly across occupations. Stanford and ADP’s indicators flag notably weaker growth for young workers in roles the dashboard flags as having high generative‑AI exposure — categories that include some administrative, routine sales and junior technical positions. Researchers caution, however, that exposure does not equal immediate displacement; they describe AI exposure as one of several structural forces reshaping early careers.

The gender result is nuanced. Raw data show women aged 22–25 have trailed men in employment growth, but the authors attribute much of that gap to “occupational sorting” — the tendency for women and men to cluster in different entry occupations, some of which face higher AI exposure. That distinction matters for employers and regulators: a headline gender gap can look like direct disparate impact, but occupational composition implies different remedies, from targeted reskilling to changes in recruitment pipelines.

For HR teams and unions, the dashboard furnishes a sharper view of where to prioritise interventions. Employers told to expect more scrutiny will have to weigh how hiring freezes or altered entry‑level job designs intersect with diversity goals and bargaining talks. Regulators tracking layoffs and disparate impacts will likely use such real‑time indicators to flag sectors and demographic groups for closer review.

The update does not answer every question. The dashboard’s documentation does not fully disclose all methodological choices behind the AI‑exposure metric, nor does it publish exhaustive demographic breakouts beyond age and gender that would allow independent audit of intersectional effects. The researchers also stop short of attributing causation; they identify correlations and plausible mechanisms but have not released longitudinal causal estimates that separate AI adoption from contemporaneous economic or policy shifts.

That partial transparency will matter as employers rely on these signals to make hiring and retraining decisions. Firms that use people‑analytics models to allocate entry roles, training budgets or headcount will need to combine these indicators with internal data and independent bias testing to avoid unintended disparate outcomes. For policymakers, the Canaries update offers an early warning system: it highlights where entry‑level demand is softening and where targeted workforce policy or oversight might be warranted.

As early‑career labour markets continue to adjust to generative AI, the Canaries dashboard underscores a central point for HR leaders: the technology is reshaping demand across occupations rather than uniformly displacing a single cohort. That means responses will have to be equally differentiated — from redesigning entry roles and investing in reskilling to rethinking recruitment channels to preserve both productivity and equitable access to early‑career opportunities.