Can we predict the jobs robots will do?

This report develops a task-level robot exposure index using O*NET task descriptions and Claude-assisted evidence to assess where present-day robots can perform work; it finds robots can do about 74% of physical tasks (34% of working hours) but are cost-competitive for only 0.3% of work today. The authors classify exposure by the environment required (E0–E3), validate the measure against historical labor-market outcomes, identify major barriers to adoption (manipulation capabilities, cost, regulation, and human preferences), and model scenarios showing slow broad automation absent rapid declines in robot costs or large capability gains.