Can we predict the jobs robots will do?
Anthropic releases a robot exposure index measuring how many US job tasks current robots can perform. The report finds robots can handle 74% of physical tasks, covering 34% of work hours, but are cost‑competitive for only 0.3% of tasks. It highlights gender, education and pay disparities in exposed jobs.

Anthropic researchers released a new index showing that current robots can handle seventy-four percent of physical tasks in the United States. These capable machines cover thirty-four percent of working hours across the economy. However, the report states that machines are cost-competitive for just zero point three percent of tasks. The creators used Claude to evaluate approximately nineteen thousand job tasks from a standard database to see how well machines handle physical requirements. They defined exposure based on the environmental control needed for a machine to operate. A historical analysis spanning fifty years was also conducted to validate the framework. The authors note that predicting the exact pace of future advances remains difficult. Factors such as regulations, human preferences, and technical capabilities pose unknown barriers to adoption. Additionally, the index does not account for technologies still in development or how shifting expenses might alter future integration.