Anthropic says its workforce has been getting older, but is also spending more time in meetings
Credit: Anthropic
Anthropic says it is hiring fewer people for junior and entry-level jobs.
Many technical workers are becoming managers of AI agents, so they are spending more time in meetings.
The company says its heavy use of AI makes it a useful early warning system for understanding how AI could affect jobs.
Anthropic, one of the world's biggest AI developers, has published a new study about how AI is changing jobs inside the company. But it is not yet clear whether these changes are good or bad news.
One worrying finding is that the share of junior workers has fallen sharply. By the middle of 2026, junior workers made up just 14% of Anthropic's technical staff, down from 21% in early 2025.
The share of new technical hires who were junior workers fell even more during the same period, from 27% to just 9%.
Is AI replacing some workers, but not all?
The study says AI is doing more work that was once handled by entry-level employees. As a result, Anthropic has not needed to hire as many junior workers.
Overall, the company also seems to be hiring fewer people because AI is helping its employees get more work done. Anthropic explained that it has hired fewer people for each dollar of revenue than large technology companies did in an earlier generation.
Importantly, Anthropic is not saying that the number of technical workers is falling. Instead, its technical workforce is still growing, but these employees are taking on more advanced work and jobs that involve managing others.
One of AI's biggest effects at Anthropic has been on how much code each engineer produces. The number of code changes submitted by each engineer rose 5.1 times from the end of 2025 to the middle of 2026. Other measures of work also increased: data queries per data scientist rose 3.1 times, contracts per legal worker rose 3.1 times, and deals per salesperson rose 1.7 times.
Humans remain in the loop
The company also found that many employees are moving into jobs that involve making decisions. As a result, they are spending more time in meetings. Across the workforce, the middle value for meeting time increased by 20%.
Engineers said they were becoming leaders and managers of AI agents. Employees in non-technical jobs said they were spending more and more time on the planning and decision-making parts of their work.
Even though automation has helped employees get much more done, some tasks still need people. These include sales calls and job interviews.
However, people can also slow down AI work because AI systems sometimes have to wait for them. The study found that Claude spends about 80% of the time in interactive sessions waiting for a human to respond.
Anthropic has also considered how work might change across other companies in the future. One major change could be in hiring, as employers may put more value on skills such as working independently and making good decisions.
There is also a more hopeful way to look at the fall in hiring: the number of job opportunities could still grow quickly. Anthropic suggested that one possible future could involve start-ups run by a single person, with a human founder directing large groups of AI agents to generate billions of dollars in revenue.
Anthropic says its heavy use of AI makes it a useful example for studying these changes. It describes itself as an early warning system, suggesting that looking at AI companies could help us understand how AI may affect jobs more widely.
