Young worker hiring falls 28 percent in AI-exposed roles
BDF Decoder

Young worker hiring falls 28 percent in AI-exposed roles

Permanent contract hires for 21-to-29-year-olds in occupations most exposed to generative AI fell 28 percent by March 2026, Banque de France research shows. The adjustment occurs through reduced recruitment and lower turnover rather than net job cuts, leaving overall employment stable.

Entry gates close first

Banque de France researchers Marco Guido Palladino, Antonin Bergeaud and colleagues tracked workforce movements between January 2019 and March 2026.

Permanent contract hires for workers aged 21 to 29 in the quintile most exposed to generative AI dropped 28 percent below November 2022 levels, compared to an 8 percent decline in the least exposed quintile.

Contract terminations also fell by 23 percent versus 6 percent, pointing to sharply reduced turnover.

For workers aged 50 and older, exposure created virtually no hiring gap.

Across roughly 6,000 tracked employers, 600 intensive AI adopters saw junior hiring fall without showing any statistically significant drop in overall headcounts.

Overlapping shocks and frozen turnover

Strict dismissal rules in France push firms to adjust staffing via hiring freezes rather than redundancies.

Yet multiple macroeconomic developments cloud direct causality.

The post-pandemic tech correction, tighter monetary policy, and remote work expanded concurrently, targeting the exact occupations most exposed to AI.

International findings also diverge: while United States data shows similar drops in junior hiring, Norwegian and Danish studies observe task reorganization without employment contractions.

A quiet blow to career ladders

Tracking aggregate jobless rates misses the stealth restructuring underway in skilled labor.

By curbing junior recruitment instead of dismissing incumbents, employers preserve current stability while severing early career ladders.

Sacrificing entry-level training risks permanently eroding human capital for marginal software efficiency.

Report an error