Anthropic reveals the jobs AI will likely replace

New research from anthropic reveals how artificial intelligence is really affecting real-world jobs. The company analyzed the usage data of its AI Assistant Claude and found that while some professions are experiencing increasing automation, many jobs remain largely intact.

During testing, each job received a coverage score. A high score meant the AI ​​is already performing a measurable portion of that role’s tasks in practice, a low score meant the job is less likely to be replaced.

The gap between AI capability and its actual use in the real world

According to research, there is a big difference between what AI could theoretically automate and what companies currently use it for. For example, AI systems could handle up to 90% of tasks in administrative functions. However, real-world usage data shows that adoption is still much lower.

Even in technology-related fields (the sector most affected by AI so far), observed AI use covers approximately one-third of job tasks. This suggests that large-scale job replacement has not yet occurred, despite rapid advances in artificial intelligence tools.

Jobs that are already under pressure from AI

Computer programmers show 75% task coverage, the highest in the data set. According to research, the use of Claude in coding leans towards full automation rather than productivity assistance.

Customer service representatives come in second place. The core tasks in that role increasingly appear in first-party API traffic, which researchers describe as companies routing work through AI channels rather than human agents.

They are followed by data entry workers with a coverage of 67%. Financial analysts and office administrators also appear among high-exposure occupations, although actual adoption in administrative roles still lags behind theoretical capacity.

Bureau of Labor Statistics data cited in the research shows that for every 10 percentage point increase in a job’s AI coverage score, projected employment growth for that role falls by 0.6 percentage points through 2034.

30% of American workers cannot be replaced by AI

About 30% of American workers score zero. Your tasks do not appear in the AI ​​usage data at any meaningful level.

Research identifies them as roles that depend on physical presence, sensory judgment and real-time situational reading. Zero-exposure occupations listed in the study include cooks, motorcycle mechanics, lifeguards, waiters, dishwashers, farm workers and court attorneys.

The Bureau of Labor Statistics projects steady growth for blue-collar jobs throughout the decade. According to research, healthcare is adding approximately 40,000 jobs per month, and demand for nurses, therapists and care workers outpaces the displacement of AI in those fields.

The impact of AI may affect administrative roles the most

A surprising finding is that workers most exposed to AI tend to be older, highly educated, and better paid. Researchers found that employees in positions exposed to AI earn about 47% more than those in positions without exposure.

This differs from previous waves of automation, which often hit lower-wage jobs first. Instead, AI appears to be aimed at tasks commonly found in white-collar professions.

Hiring trends may already be changing

Although the research did not find a major increase in unemployment, early signs of change in hiring patterns are emerging. Among workers ages 22 to 25, the job search rate in AI-exposed occupations has dropped approximately 14% since the launch of ChatGPT in late 2022.

Researchers say this could indicate that companies are slowing hiring in roles where AI tools can already help with key tasks.

What the findings mean for people

The data suggests that AI is reshaping tasks rather than eliminating entire professions. Roles that rely on physical skills, human interaction, or real-world environments seem much more resilient.

At the same time, office professions that rely heavily on digital workflows may continue to face increasing automation pressure as AI tools become more capable.

Anthropic describes the index as a first step and plans to update coverage measures as usage data changes.