Fewer juniors, more agents: Anthropic's internal data on how work is changing

On 16 October 2026 the Economics team of the Anthropic Institute published the study "How employment is changing at Anthropic", one of the first inside looks at how deploying AI agents reshapes the workforce of an AI lab. According to the document, the share of junior employees in the company's technical teams fell from 21% at the start of 2025 to 14%, while juniors as a share of new technical hires dropped from 27% to 9%, alongside a sharp decline in entry-level job postings after December 2025. Over the past year, technical staff as a share of total headcount also fell from 61% to 56%, in a company that tripled in size over the same year: the report does not mention layoffs. Juniors make up a smaller proportion, but they were not cut. In April 2026, roughly a third of machine learning managers and a fifth of software engineering managers said they were holding back on hiring because of automation.

The organisation's telemetry shows that in March 2026 the total working hours of autonomous and interactive agents exceeded human working hours for the first time; by August, agents directed by employees were working six times as many hours as the people supervising them. About 80% of the time in interactive sessions is spent waiting for human instructions, approvals or feedback. Code commits per engineer rose roughly fivefold compared with six months earlier, while coding sessions started by non-technical staff climbed from 2 a month in August 2025 to 90 in May 2026. To measure productivity, the study compares 192 engineers who adopted coding agents with 74 who did not; the adopting group, however, was already on a steeper output trajectory. For agent-written code, measured through deletions and reverts, there is no quality difference compared with code written by hand by the same person.

The scorecards used in job interviews show recruiters paying more attention to candidates' autonomy and judgement, at the expense of execution and specialist expertise. The company has announced an early-warning system to track roles undergoing change, along with priority for internal redeployment. The study's authors themselves acknowledge, however, that these measures do nothing about the reduced hiring of juniors, and they call for public policies to support and retrain workers. The report is published by Anthropic itself, covers a single company (the authors say it is not fully replicable elsewhere), relies partly on anecdotes from interviews and uses survey questions that changed over time. The main figures were reported independently by Business Today on 19 October.

The shrinking number of entry-level positions does not come from headcount cuts; some managers explicitly link it to AI. If frontier labs hire fewer people at the start of their careers, it remains unclear where the next generation of engineers will build the hands-on experience and critical judgement that hiring processes now value most. — Olya