Anthropic internal study: junior share of technical staff fell from 21% to 14% as AI agents log 6x more work hours than humans
Anthropic published an internal study reporting that the junior share of positions in its technical departments fell from 21% to 14% since early 2025, and the junior share of new technical hires fell from 27% to 9% (open junior roles at 6%), with the sharpest decline from December 2025. AI agents have logged more hours than human employees since March 2026, reaching a 6:1 company-wide ratio by August 2026 (19:1 in research). Reported productivity gains: engineers 5.1x code commits, data scientists 3.1x queries, legal 3.1x contracts, sales 1.7x deals, recruiters 1.4x hires, accounting 1.4x invoices. Non-technical departments stayed mostly stable.
Entities: Anthropic, Claude, Nolan Lovett, NATO Special Operations University, The Decoder
0 primary
What happened
Anthropic published an internal study saying the junior share of its technical positions fell from 21% to 14% since early 2025. The junior share of new technical hires fell from 27% to 9%, with the sharpest drop from December 2025. The study also says AI agents have logged more hours than human staff since March 2026, reaching 6:1 company-wide by August 2026 and 19:1 in research. Reported productivity multiples include 5.1x code commits for engineers and 1.4x invoices for accounting. Non-technical departments stayed mostly stable. The coverage links no study, so the figures are unverified.
Why it matters
This is one of the few measured, company-level data points on AI and entry-level work, which makes it useful to hiring managers, graduates and workforce planners. It suggests junior technical roles may be the first to shrink where agents are heavily used. The evidence is weak for wider conclusions. It comes from one AI-heavy vendor with fast growth and voluntary adoption, and Anthropic itself says the results may not generalise. Treat it as a signal worth testing elsewhere, not as a trend that has been shown.
What is noise
The headline "nearly stopped hiring" goes further than the data. Junior share of hires fell to 9%, which is not zero, and the total number of hires is not given. "Six times as many hours" compares agent compute time with human working hours, which does not mean six times the output. The "tragedy of the cognitive commons" and the 2030 to 2045 expert-shortage argument are speculation layered on top. The productivity multiples are self-reported by the vendor, with no method or independent check.
Watch next
- 01Whether Anthropic releases the full study with methodology and underlying numbers, including absolute headcount and hiring totals, so the percentages can be checked.
- 02Independent data from non-AI firms, such as UK or US graduate and entry-level tech hiring figures, and company disclosures showing a similar fall in junior hiring over the next two to three quarters.
- 03Whether Anthropic's junior share of technical hires stays near 9% or recovers in later updates, and whether other AI labs publish comparable workforce figures.
Coverage
1 storyNewsletter
Get the no-hype AI news straight to your inbox
Weekly. No spam, unsubscribe any time. Privacy
More labour signals
Full feed →- NYT-led publishers file summary judgment brief citing internal OpenAI/Microsoft messages calling AI training "astonishing theft" and admitting chatbots substitute for journalism18 Sept 202686
- Character.AI files for Chapter 11 bankruptcy and proposes selling assets to Disney as lead bidder for $20 million13 Oct 202677
- Character.AI files for Chapter 11 bankruptcy as part of a deal letting Disney hire most staff and bid for a perpetual license to its technology, leaving liabilities behind19 Oct 202676
- Unsealed filings in NYT v. OpenAI/Microsoft copyright suit reveal internal admissions calling AI training data scraping "theft" and an "existential threat" to publishers17 Sept 202674