In a study of internal data and 17 employee interviews, the company says junior-level employees fell from 21% to 14% of its technical departments since early 2025, and junior technical hires dropped from 27% to 9% of hires.
The report says that by August, interactive AI agents, meaning software that carries out tasks on an employee's behalf, logged six times as many working hours as the company's human staff.
It says output rose as employees adopted coding agents: code commits per engineer reached five times their level of six months earlier, and lawyers, accounts payable staff and salespeople also handled more work.
To separate AI's effect from the company's rapid business growth, the authors followed 192 engineers who adopted coding agents and compared them with 74 who did not, finding sharp increases in code added.
They found no difference in how often agent-written code was later removed or reverted, though they caution that tool adoption was not random and that more code is not necessarily better code.
Anthropic also says it takes in two to four times as much revenue per employee as Meta and Google did at comparable points in their growth.
The study says human attention has become the bottleneck: interactive Claude sessions spend about 80% of their time waiting for people, and reading all the output would take even non-technical staff over half a work week. Employees increasingly do "meta-work" such as writing instructions for agents and monitoring pipelines, which rose from 10% to 22% of interactive sessions. Hiring managers now stress autonomy, judgment and creative AI use over narrow technical expertise. Anthropic says it has built an early-warning system for jobs at risk and prioritizes moving employees internally, but acknowledges these steps do not address the slowdown in junior hiring and calls for government retraining and income-support policy.