Insights ¦ Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence

Key Take Aways

1. Entry-Level Impact: The proliferation of generative AI is beginning to affect entry-level employment, particularly for workers aged 22–25 in highly AI-exposed occupations.
2. Employment Decline: Early-career workers in the most AI-exposed roles have experienced a 13% employment decline compared to pre-adoption levels, even after accounting for firm-level shocks.
3. Cohort Disparity: While overall employment remains strong, young workers in AI-exposed jobs have seen stagnant employment since late 2022, unlike older workers who continue to experience growth.
4. Automation vs Augmentation: Declines are concentrated in roles where AI automates rather than complements tasks; jobs with more augmentative AI use have stable or increasing employment.
5. Direct AI Link: Employment reductions persist even after adjusting for firm-specific shocks, suggesting a direct relationship between AI adoption and workforce decline.
6. Wage Stickiness: Wages have not declined as noticeably as employment, indicating stickiness in pay despite shifting job patterns.
7. Consistent Patterns: Effects are consistent across AI exposure measures (e.g. GPT-4, Claude queries) and robustness checks (e.g. excluding tech sectors or cropping samples).
8. Intrinsic to AI: Industries unaffected by outsourcing or remote work also show employment drops, implying AI’s automation potential, not labour market trends, is the cause.
9. Disproportionate Effect: More experienced workers and those in less AI-exposed occupations have been less affected, indicating a disproportionate impact on less-skilled, younger staff.
10. Codified Knowledge Displacement: The disruption is likely tied to AI’s ability to replace codified knowledge—typically a larger part of young workers’ roles.
11. Career Progression Risks: Slower employment growth for young workers may impact long-term income and career development trajectories.
12. Ongoing Monitoring Needed: These early disruptions warrant continuous monitoring to assess broader structural changes in the labour market.

Key Statistics

• Employment for 22–25-year-olds in AI-exposed occupations declined by nearly 20% since late 2022.
• Young workers in the top AI exposure quintile saw a 6% employment drop, while older workers in the same group saw a 6–9% rise.
• A 13% relative decline in employment for young workers in the most AI-exposed roles was found even after controlling for firm-level effects.
• Employment overall has grown post-pandemic, but entry-level jobs in exposed roles have plateaued since late 2022.
• Declines are driven by automation use, aligning with task substitution models.
• Wage changes remain minimal across age and exposure groups.
• Results hold across education levels and beyond the tech sector.
• Impacts were also seen in non-remote and non-outsourced occupations.
• Employment dropped significantly among software developers, customer service representatives and health aides in high AI exposure groups.
• The use of fixed-effects models still shows a decline even when controlling for firm/industry factors.
• Pre-GPT-4 (pre-2022) data shows no such divergence, reinforcing the link to recent generative AI expansion.
• Similar trends are observed across gender and employment type (part-time or temporary).

Key Discussion Points

• Generative AI adoption is threatening entry-level employment, especially for young people in high-exposure roles.
• Vulnerability stems from AI’s ability to replace codified knowledge, often the basis of younger workers’ job functions.
• Despite a robust job market overall, the stagnation in youth employment may signal long-term risks.
• Task type matters—automative AI affects roles more negatively than augmentative applications.
• Declines are not explained by industry trends or outsourcing but appear tied to AI adoption.
• Wage resilience suggests rigidities in salary structures even as roles are displaced.
• Robust findings across test models confirm the structural nature of the changes.
• The disruption coincides with post-2022 proliferation of generative AI tools like GPT-4.
• Education level does not insulate workers from impact; both highly and less-educated groups are affected.
• Vigilance is necessary as early patterns may signal deeper labour shifts ahead.
• The findings fuel ongoing debates on AI’s role in augmenting versus replacing human work.
• Policymakers and businesses must respond to early indicators of labour reallocation, particularly to protect young and entry-level workers.

Document Description

This article examines the emerging impact of generative AI on the U.S. labour market, focusing on entry-level workers in AI-exposed roles. Drawing on high-frequency payroll data from the largest U.S. provider, the study identifies six key findings that reveal a disproportionate employment decline among younger workers since late 2022. The research differentiates between AI automation and augmentation, tests its conclusions against multiple robustness checks, and finds a direct link between AI exposure and employment outcomes. The paper provides early but significant insights for policymakers, executives and workforce strategists navigating AI’s growing influence on job markets.