Anthropic released what it calls the Anthropic Economic Index on January 15, a framework for tracking which occupations are most exposed to displacement by large language models. The paper, authored by economists Maxim Massenkoff and Peter McCrory, introduces a metric called “observed exposure” that compares AI’s theoretical task capability against actual usage data drawn from anonymized Claude interactions in professional settings.
Computer programmers top the exposure list at 74.5% observed task coverage, followed by customer service representatives at 70.1% and data entry keyers at 67.1%. Medical record specialists, market research analysts, and financial analysts rank among the next most affected. Workers in the most exposed occupations are 16 percentage points more likely to be female, earn 47% more on average, and are nearly four times as likely to hold a graduate degree compared to those in the least exposed jobs. The profile is the lawyer, the financial analyst, the software developer, not the warehouse worker.
The theoretical ceiling dwarfs current adoption. In computer and mathematical occupations, large language models could handle an estimated 94% of tasks, but Claude currently performs roughly 33% in observed professional use. Office and administrative roles show a similar disparity.
Roughly 30% of occupations register zero AI exposure in the index, jobs requiring physical presence that no language model can replicate: cooks, mechanics, bartenders, dishwashers. Despite the high theoretical exposure for white-collar roles, unemployment rates among the most affected occupations have not risen at statistically meaningful rates since ChatGPT’s release in late 2022. The researchers describe the average change as “small and insignificant.”
Where the data does shift is in hiring. Workers aged 22 to 25 in exposed occupations have experienced a 14% drop in the job-finding rate compared to 2022, though the researchers note that finding is just barely statistically significant. A separate study found a 16% fall in employment in AI-exposed jobs among workers in the same age bracket.
Massenkoff said the tool is designed to detect disruption before it becomes obvious in aggregate statistics, comparing the challenge to tracking the “China shock” of the early 2000s, where major employment effects took years to surface clearly in the data.
The paper names the scenario its framework is built to detect: a “Great Recession for white-collar workers,” in which unemployment in the top quartile of AI-exposed occupations doubles from 3 to 6%. That has not happened. The researchers say their index would identify it if it did.
