Anthropic's Own Numbers Show Junior Hiring in Freefall
Anthropic has put a number on what AI is doing to its own hiring funnel, and the number is stark: junior-level hires in its technical departments fell from 27% of new hires in early 2025 to just 9% today. Open junior technical roles have shrunk further, to roughly 6% of postings, with the steepest drop beginning in December 2025. It lands on an industry already nervous about an AI-driven freeze on entry-level hiring, giving that anxiety its most detailed numbers yet from inside a frontier lab — and it is the lab itself publishing the evidence.
The figures come from a report Anthropic released this week, "How employment is changing at Anthropic," built on internal hiring records, calendar data, and an April 2026 employee survey. The company says its own AI agents have logged more working hours than its human staff since March 2026; by August the ratio had reached 6-to-1 company-wide and 19-to-1 inside the research division. Technical departments' share of total headcount slipped from 61% to 56% over the past year even as headcount grew. Revenue per employee at the five-year mark is now 2 to 4 times what Meta and Google posted at the same stage, by Anthropic's own comparison, and employees with at least a year of tenure were 50% more likely to switch departments each month in 2026 than in 2025 — a workforce growing less by adding people than by moving the people it already has.
The entry-level cliff
The contraction isn't limited to new hires. The share of junior employees already on technical teams fell from 21% to 14% since early 2025, and the strain shows up in how the remaining staff work, not just who gets hired. The median employee now spends 20% more time in meetings than a year ago; technical individual contributors are up 16%, which Anthropic attributes to a growing need for human coordination as agents absorb execution work. On output, the company's own accounting shows engineers committing roughly 5x the code of six months ago, legal processing 3.1x the contracts, and sales reps closing 1.7x more deals. Non-technical employees' use of coding agents exploded too, from about two sessions a month in August 2025 to 90 by May 2026 — evidence that the line between "technical" and "non-technical" work, and who still needs a junior technical hire at all, is dissolving.
A warning Anthropic published about itself
The report doesn't read as pure triumph. It cites outside researcher Nolan Lovett of NATO's Special Operations University, who calls the risk a "tragedy of the cognitive commons": companies that swap entry-level roles for agents capture the efficiency now, while the cost of a thinning expert pipeline is paid later, by everyone who eventually needs a seasoned specialist. Lovett puts the reckoning at 2030 to 2045, since today's senior practitioners were trained five to twenty years ago in roles that increasingly don't exist for the next cohort; he flags software engineering, financial analysis, and legal research as the most exposed fields. Separately, about a third of Anthropic's ML managers and a fifth of its software engineering managers told the company they are hiring less specifically because of AI.
"A workforce that is anxiously worried is not a workforce doing its best work." — from Anthropic's report
Anthropic hedges its own findings, noting it is an unusually AI-saturated software company and that the pattern may not generalize to the rest of the economy. But for anyone tracking AI hiring trends, the relevant fact isn't whether Anthropic is typical — it's that the company building the agents responsible for the shift is the one quantifying junior hiring's collapse, in public, eighteen months after it started. Decision-makers watching entry-level pipelines at their own AI-adopting firms, and investors pricing labor-cost savings into AI valuations, now have a frontier lab's own numbers to benchmark against — and a named researcher's warning about what that bargain costs a decade from now.