Anthropic, the company behind the Claude chatbot, published a study of its own workforce on Oct. 16 that says AI is changing whom it hires. By the company's count, junior roles fell from 27% of its technical hires to 9%, and junior staff slipped from 21% to 14% of its technical departments since early 2025. Anthropic's technical hiring is still growing in total, and it says it has tripled in size in the last year, so the shift is in who it hires while its workforce keeps growing. In an April survey, about a third of its machine learning managers and a fifth of its software engineering managers said they were hiring less because of AI.
The study ties that shift to AI agents, AI that can carry out tasks on its own, such as writing and fixing code, instead of only answering questions. Anthropic says that by August the agents its employees direct were running six times as many hours as the employees worked, counting each person at 40 hours a week and adding up the helper agents each session starts. It says engineers are saving code changes five times as often as six months earlier, and that when it followed 192 engineers who took up coding agents and 74 who did not, the adopters added sharply more code in the months after they started, compared with the others. Lawyers handled more contracts and salespeople closed more deals as the tools spread, it says.
People have become the slow part of the system, the report says. In sessions where a person works back and forth with them, agents spend about 80% of the time waiting for that person to respond. Even for staff outside technical teams, reading everything their agents produce would take at least half the work week, by the company's estimate. For technical jobs, hiring managers now put more weight on a candidate's independence and judgment, and less on expertise in one field, according to the company's review of its interview notes.
The numbers come from Anthropic's own records and staff interviews. Anthropic says its business boomed over the same period, which also lifts output, and that engineers chose for themselves when to start using the tools, so those gearing up for a big project may have started first, which could skew its estimates. It also says more code does not necessarily mean better code. It found that code written by agents was no more likely to be removed or undone later than code the same engineers wrote by hand, while noting that code quality has many sides and that AI's mistakes may be harder to spot.
Anthropic says it has set up an early warning system to track jobs at risk from AI and is moving affected employees into other roles: staff who had been there at least a year were 50% more likely to move between departments each month in 2026 than in 2025. It warns that the same patterns, spread across the economy, suggest AI could put workers out of jobs, and it says none of its own steps directly address the slowdown in junior hiring, since protections for current staff do not help people who were never hired. The report does not say whether any Anthropic employee lost a job, or how much of the trend other companies will see.