The jobs most affected by AI are concentrated in white-collar work, according to a new Anthropic study that ranked U.S. occupations by how much of their work artificial intelligence can now help do or automate in practice. Computer programmers led the list, followed by customer service representatives and data entry keyers.
Anthropic, the company behind the Claude chatbot, said its new measure tracks both what AI can do in theory and how workers are actually using it on the job.
Researchers said that, so far, they see limited evidence that AI has had a clear effect on overall employment. Still, the study found early signs that hiring may be slowing for younger workers in more exposed occupations.
Anthropic reports 10 jobs most affected by AI
The top 10 most affected occupations in the study were:
1. Computer programmers (74.5%)
2. Customer service representatives (70.1%)
3. Data entry keyers (67.1%)
4. Medical record specialists (66.7%)
5. Market research analysts and marketing specialists (64.8%)
6. Sales representatives, wholesale and manufacturing, except technical and scientific products (62.8%)
7. Financial and investment analysts (57.2%)
8. Software quality assurance analysts and testers (51.9%)
9. Information security analysts (48.6%)
10. Computer user support specialists (46.8%)
Anthropic’s Revealing Chart on AI’s Impact on Jobs
Anthropic has unveiled a pivotal chart that underscores the chasm between AI’s capabilities and its real-world application in the workforce.
Derived from analyzing 2 million actual conversations with Claude, this radar chart,… pic.twitter.com/eu8E7D1zlL
— Brian Roemmele (@BrianRoemmele) March 6, 2026
Researchers said exposure does not mean an entire job can be replaced. It means a share of a job’s tasks could be sped up or handled by AI. Anthropic said jobs combine many tasks, and some remain hard to automate. A teacher, for example, may use AI to grade work, but not to run a classroom.
Ranking was built using task and AI usage data
To build the ranking, the company combined occupational task data from O*NET, Claude usage data and earlier estimates of which tasks large language models could do faster. It gave more weight to work-related and automated uses than to uses where AI only assists a worker.
Anthropic said actual AI use still trails what the technology could theoretically do. Legal rules, software limits and human review can slow adoption.
The study also found that occupations with higher exposure are projected by the U.S. Bureau of Labor Statistics to grow more slowly through 2034. Workers in the most exposed jobs were more likely to be older, female, more educated and higher paid.
The jobs most affected by AI stood apart from lower-exposure roles that rely more on physical work. Anthropic said jobs with little or no observed exposure included cooks, bartenders, lifeguards, groundskeepers and motorcycle mechanics.
On the labor market, researchers said they found no clear rise in unemployment for highly exposed workers since late 2022. But they said the data suggests hiring into exposed jobs may have slowed for workers ages 22 to 25.
