OECD measures AI capability gaps across occupations
Exposure rankings have limits as employment evidence.
- Event / announcement date
- 2026-05-26
The OECD’s May 26 paper maps nine cognitive, social and physical AI capabilities against job requirements. Analysts rated 40 occupations; four language-model families extended this to 879. Its July revision still sought validation by occupational experts, workers and employers. Smaller gaps indicate greater potential exposure. [1]
The ILO’s separate May 2025 explanation takes a task-based route, combining task scores and their variation within occupations. It stresses that exposure describes technical potential, while electricity, broadband, skills and costs affect adoption. A score cannot establish that a particular employer installed a system or eliminated a position. [2]
One total can hide differences in language, social judgement and manipulation, the OECD cautions. Exposure measures capability proximity, not adoption or job losses. [1]
The source
- OECD: AI exposure measure ↗
Source date 2026-05-26 · reviewed October 29, 2026 · PDF revised July 2026. - ILO: Occupational exposure ↗
Source date 2025-05-20 · reviewed October 29, 2026
About this item: Retrospective report. No first-hand reporting or independent testing is claimed. Corrections policy.
