Good morning. Anthropic’s recent study mapping AI’s reach across hundreds of occupations continues to raise fresh concerns about the future of white-collar work.
Economists Maxim Massenkoff and Peter McCrory analyzed millions of real Claude conversations, matched them against 800 occupations, and found a striking gap: AI can theoretically automate 94% of computer and math tasks but currently handles only about 33%. In business, finance, legal, and office administration roles, the story is similar. Financial and investment analysts are specifically identified as one of the most “exposed” roles.
For more insight on what’s behind the research, my Fortune colleague Matt Heimer sat down with McCrory, head of economics at Anthropic. He makes the case that exposure data could help corporate leaders, policymakers, and individual professionals adapt their workflows and careers to AI and perhaps help head off severe job-market disruptions before they become major social problems.
What makes this study different from the usual AI-disruption commentary is the data source—actual Claude usage data from the workplace. It sheds light on what share of the work in a given occupation AI systems can already do, and how much more they could theoretically take on.
For example, for business and finance occupations broadly, the theoretical exposure (tasks AI could speed up by more than 50%) is very high, but actual observed usage still lags behind. That means the potential disruption hasn’t been fully realized. That gap, however, is expected to close as capabilities improve and adoption deepens.
In McCrory’s conversation with Heimer, he goes into depth on the implications of the study findings, including, as an economist himself, what the degree of current exposure versus theoretical exposure looks like for his own work. You can read the interview here.
Sheryl Estrada
sheryl.estrada@fortune.com
