Stanford study finds AI hits entry-level jobs hardest, closing the career on-ramp

A Stanford study has found that AI's effect on the labor market is concentrated at the bottom of the career ladder: entry-level positions are losing ground fastest, particularly in occupations built on knowledge that can be formally taught and written down. According to Ars Technica, which reported on the research, the pattern points toward a future in which incumbent workers largely keep their jobs while the openings that once absorbed new entrants quietly disappear.

Codified versus tacit knowledge

The researchers' central explanation rests on a distinction between two kinds of workplace knowledge. Entry-level roles, they argue, tend to run on codified knowledge — material that is standardized, documented and teachable through schooling, textbooks or written procedures. Senior roles instead draw on tacit knowledge, the kind that accumulates through practice, mentorship and repeated exposure to real situations, and which AI appears to complement rather than replace.

To test the idea, the team used the formal education requirements recorded in the O*NET occupational database as a proxy for how dependent each occupation is on codified knowledge, then compared that against employment data. The result, per Ars Technica: occupations with a heavier reliance on codified knowledge showed slower entry-level employment growth, while occupations that lean on tacit knowledge saw employment grow faster for mid-career and senior workers.

Education as a buffer

The data also suggested that a highly educated workforce softens the blow. In occupations where a larger share of workers hold college degrees, the employment gap between AI-exposed and less-exposed roles was noticeably smaller. Where few workers were graduates, the split was starker — the least exposed occupations kept adding jobs while the most exposed ones were declining in employment outright, according to the study.

A warning from the lead researcher

Lead researcher Erik Brynjolfsson told The Washington Post that the trends imply a near future in which people who held jobs before AI largely retain them, while many of the positions that once took in the incoming working-age cohort disappear. "The entry-level effects we're measuring are real, persistent and widening," he said, adding that he has grown more concerned about a labor market that sustains its overall employment level while quietly shutting the on-ramp for people at the start of their careers.

Why it matters

For the tech industry, the findings land close to home. Entry-level software work is largely structured around exactly the kind of codified knowledge the study describes — documented procedures, standard toolchains, skills learnable from courses and manuals — which places junior developer roles squarely in the exposure zone the researchers identified. If the pattern holds, companies may retain their mid-career and senior engineers while the traditional first rung of the profession narrows.

That creates a longer-term problem the study's own framework makes explicit: the tacit knowledge that still favors senior workers is acquired precisely by holding those early jobs. Squeeze the on-ramp, and the pipeline that eventually produces experienced practitioners thins out. For new graduates, educators and hiring managers, the research reframes AI's labor-market impact not as headline unemployment but as a blocked entrance — one that aggregate employment statistics will never reveal.