Mar 5, 2026Economic ResearchLabor market impacts of AI: A new measure and early evidence

2026-02-26 · Source: Anthropic Research · Field: Finance & Economics — Economic Analysis & Policy, Artificial Intelligence & Machine Learning · Depth: Advanced, extended

Summary

A new measure of AI displacement risk, called "observed exposure," has been introduced, combining theoretical Large Language Model (LLM) capability with real-world usage data, emphasizing automated and work-related applications. This measure, detailed in a March 5, 2026 report, reveals that AI's actual coverage in professional settings remains a fraction of its theoretical potential. Occupations with higher observed exposure, such as Computer Programmers (75% coverage), Customer Service Representatives, and Data Entry Keyers (67% coverage), are projected by the BLS to experience slower growth through 2034. Workers in these highly exposed professions tend to be older, female, more educated, and higher-paid. While no systematic increase in unemployment for highly exposed workers has been observed since late 2022, there is suggestive evidence of a slowdown in hiring for younger workers (ages 22-25) in these occupations.

Key takeaway

For AI Scientists and Research Scientists evaluating AI's labor market effects, your analysis should prioritize the "observed exposure" metric, which combines theoretical LLM capability with actual, automated usage data. This approach offers a more nuanced view than theoretical potential alone, helping to identify jobs most susceptible to AI-driven changes. Pay close attention to hiring trends for younger workers in highly exposed occupations, as this may be an early indicator of AI's impact on labor market entry, even if aggregate unemployment remains stable.

Key insights

AI's labor market impact is better understood by combining theoretical capability with observed, automated usage.

Principles

• AI's actual usage lags theoretical capability.
• Observed exposure correlates with slower job growth.
• Unemployment impacts are not yet systematic.

Method

The "observed exposure" metric integrates O*NET task data, Anthropic Economic Index usage, and theoretical LLM task exposure (β) from Eloundou et al. (2023), weighting automated and work-related uses more heavily.

In practice

• Monitor "observed exposure" for job impact signals.
• Analyze hiring rates for young workers in exposed roles.
• Focus on unemployment as a key harm indicator.

Topics

• AI Labor Market Impact
• Large Language Models
• Occupational Exposure
• Employment Projections
• Anthropic Economic Index

Best for: AI Scientist, Research Scientist, AI Researcher, Data Scientist, Policy Maker

Related on AIssential

• Anthropic Says AI is Not “Killing Jobs”, Shares New Way to Measure AI Job Impact — Actual AI job impact is measured by "Observed Exposure," tracking real usage over theoretical capability.
• Apr 22, 2026Economic ResearchWhat 81,000 people told us about the economics of AI — AI exposure correlates with job displacement concerns, while productivity gains vary across income and career stages.
• AI’s effect on labor productivity is murkier than you might think - marketplace.org — AI's impact on economy-wide labor productivity remains uncertain and difficult to measure, despite clear firm-level benefits.
• Apr 22, 2026Economic ResearchAnnouncing the Anthropic Economic Index Survey — Qualitative data from AI users can reveal economic shifts faster than traditional metrics.
• How Human Work Will Remain Valuable in an AI World — AI's real-world impact is constrained by physical, regulatory, and organizational factors, not just technological speed.
• Is AI to blame for hiring woes faced by college graduates? - ABC News - Breaking News, Latest News and Videos — AI's impact on recent college graduates' job prospects is debated, with evidence suggesting sector-specific declines since late 2022.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Anthropic Research.