Atlas AI Resources

← Back to AI News

AI research

Anthropic is hiring fewer junior tech workers as its own AI agents do more of the work

Published October 20, 2026

Anthropic, the company behind the Claude chatbot, published a study of its own workforce on October 16. Its headline finding: as AI agents (AI tools that carry out multi-step tasks on a computer, not just answer questions) took over more coding and bug-fixing, the company sharply cut the share of junior people it hires for technical jobs. It is one company’s view of itself, but it is unusually detailed, and it is about the first rung of the career ladder.

What happened

The report, “How employment is changing at Anthropic,” was written by four Anthropic researchers: Nathan Wilmers, Regina Catipon, Maxim Massenkoff and Peter McCrory. It draws on internal records on staffing, AI use, output, surveys, calendars and hiring decisions, plus 17 employee interviews.

The core numbers: junior employees fell from 21% to 14% of Anthropic’s technical departments since early 2025, and junior hires fell from 27% to 9% of technical hiring. Business Insider, reading the report’s chart, notes the drop was milder outside technical teams, where junior hires went from 31% to 26%. Technical hiring still grew in absolute numbers, because the company tripled in size over the past year; sales, finance and recruiting simply grew faster.

Anthropic also says that by August, the agents its employees directed logged six hours of runtime for every hour its staff worked. Read the fine print: that counts time agents spent computing and using tools, adds up agents running side by side, and assumes each employee works 40 hours a week. It is not six digital workers per person. The report also finds those agents spend about 80% of a session idle, waiting for a human to respond.

Output rose too. Code commits (saved batches of code changes) per engineer are five times what they were six months ago. Comparing 192 engineers who adopted coding agents with 74 who did not, the authors found output jumped after adoption. They checked whether agent-written code was later deleted or reverted more often than code the same person wrote by hand, and found no difference.

Why it matters to a normal person

Junior roles are where people learn a trade. If AI does the beginner tasks, there are fewer beginner jobs, and the report admits this. Anthropic says it built an early warning system to spot roles at risk and moves existing staff into new jobs, but the authors write that “employment protections for incumbents will not benefit those not hired to begin with.”

It also shows what employers there now look for. Hiring managers increasingly comment on autonomy and judgment rather than narrow technical skill, and mentions of a candidate’s creative use of AI tools have roughly tripled for technical roles since 2024. The human work that remains is reading, checking and deciding: the report estimates that even for non-technical staff, reading everything their Claude sessions produced would take over half a work week, and median meeting time is up 20% from a year ago. The AI generates; people still decide what is right and what ships.

What it does not mean

It is not proof about the whole job market. This is one fast-growing AI company, which sells the tools it is studying, describing itself. The report is not peer-reviewed (checked by outside experts before publication), and no one outside has audited the data.

It does not prove AI alone caused the shift. The authors say business growth made sales easier, and that the engineers who adopted agents were not chosen at random, which could inflate the effect. More code is also not the same as better products, and their quality checks are rough proxies.

It is not a layoff story, and Anthropic makes no forecast that junior hiring will hit zero. It is a clear warning sign for entry-level tech work, from a company whose CEO, Dario Amodei, has said AI may eliminate up to half of entry-level white-collar jobs in the next several years, as Business Insider notes.

Sources