How do occupations differ in their exposure to automation?

MGI's report classifies US occupations into four archetypes by how much of their work could theoretically be automated: "occupations with the lowest automation potential were classified as people-centric, while those with high shares of automatable tasks were labeled agent-centric or robot-centric," with "roles with a more even balance... grouped into mixed or hybrid archetypes." People-centric occupations — found "for example, in healthcare and in building and maintenance" — "make up about one-third of US jobs and pay an average of $71,000 a year," with "physical activity that current technologies cannot replicate" accounting for "about half of the work hours" in these roles. At the other extreme, the most automatable occupations "make up about 40 percent of total jobs," pay "an average of $70,000," and are "most... agent-centric roles in legal and administrative services" built on cognitive tasks "such as drafting documents."

The archetypes come from the report's broader automation-potential analysis (Exhibit 3), which scores occupations by the share of their task hours that current agents (nonphysical automation) or robots (physical automation) could theoretically perform, set against "the level of human proficiency required." MGI frames the taxonomy as a planning tool for leaders and workers alike: the framework "applies across labor markets and can help leaders see where change may come first and how workforce transitions could unfold," distinguishing roles likely to "evolve into human–agent–robot coworker models" from those "likely to be largely automated... under human supervision" — while stressing that even highly agent-centric work like document drafting will still need people to "guide, supervise, and verify" it.

This taxonomy sorts occupations along a different axis than What collaboration level do workers actually want with AI?, which classifies tasks by workers' preferred level of human-AI collaboration rather than by technical automation potential; MGI's agent-centric and robot-centric categories describe what machines could in theory do, not what workers want. It also differs from Why do firms build automating AI instead of pro-worker AI?, which sorts the technologies themselves by their effect on worker value — MGI instead sorts the occupations that use those technologies, by exposure. Read together with Does concentrated AI exposure enable workers to adapt and reallocate?, the near-equal pay MGI finds between people-centric and agent-centric roles (about $70,000–$71,000) is consistent with that note's finding that concentrated exposure, not mean exposure, drives labor-market outcomes.

MGI's archetypes are built from theoretical automation potential, not observed adoption or realized job change, and the report does not say how long the transition from agent-centric classification to actual automation might take for any given occupation. The pay figures are averages across broad occupational groups, so the near-equal $70,000–$71,000 averages do not establish that automation exposure is wage-neutral for individual workers within those groups, nor do they capture how pay might shift as the agent-centric and hybrid archetypes are adopted unevenly over time.