The AI Exposure Index scores 772 occupations and 1,786 academic programs across three dimensions: digital intensity, human interaction, and physical work. Built from O*NET work characteristics and BLS employment data.
Browse the index by occupation, by academic program, or read how the scores are constructed.
Explore the indexEach view shares the same three-dimensional scoring. Choose where to start: an occupation you know, an academic program you advise into, or the methodology behind the scores.
Most "AI exposure" measures collapse to a single number. This index keeps three dimensions visible, because the same job can be highly digital and also heavily human-facing, and that combination matters for what AI can and cannot do.
How much of the work happens through computers, structured data, and digital tools. Higher scores mean more of the task surface is reachable by software.
How much the role depends on real-time interaction with other people: care, persuasion, coordination, judgment in social context. Higher scores mean a stronger buffer against pure automation.
How much the work is tied to physical objects, places, and bodies. Higher scores mean the work resists software-only substitution because something has to happen in the world.
Being grounded in data is essential. In a labor market changing this fast, the decisions institutions, workforce boards, policymakers, and individuals make today carry long horizons. A program built, a training investment funded, a credential pursued: each plays out over years, into work that may not look the way the data describes it now.
The AI Exposure Index is a structured way to look ahead. It scores where work sits along three measurable dimensions, built from public O*NET data and weighted by BLS employment counts, so program decisions made today can be informed by where the work is heading rather than only where it has been.
The goal is not to predict which jobs disappear. It is to give institutions, advisors, and policymakers a frame for asking which programs prepare students for work that will keep its shape, which programs feed into work being reshaped, and where the answer depends on which dimension you weigh most.
For the full construction, including variable selection, weighting, and limitations, see the methodology.
Questions about the index, interested in collaborating, or have feedback? Send a message.