Skills in AI's path

A skill doesn't have an AI-exposure score of its own. But the jobs that use it do. So we score each in-demand skill by the AI task-overlap of the largest occupations that rely on it — employment-weighted. A high score means the skill sits in work AI can already touch; a low score means the skill tends to anchor work that today's models barely reach.

Across 92 skills, the median sits at the 68th exposure percentile. Exposure is task overlap — what AI could touch — not automation, adoption, or jobs lost. A high-exposure skill is not a worthless skill; it's one where the work is changing fastest.

How this is computed. For each skill we take the largest occupations that demand it (by employment, from the Lightcast→O*NET crosswalk) and average their AI task-exposure percentile, weighted by employment. Only skills with at least 3 matched occupations are shown. Exposure percentile is the unified task-overlap index (Eloundou GPT-overlap + Felten AIOE) ranked across all occupations. This measures where a skill sits relative to AI's current reach — a map of change, not a prediction of loss, and not a ranking of skill value.

← Back to all skills in demand