Anthropic published two reports on October 16, 2026 that use its own staff as a test case for AI at work. The first, built on internal hiring, telemetry and calendar data, finds that the share of junior staff in technical departments fell from 21% to 14% since early 2025, while AI agent hours grew to six times employee hours by August. The second, a survey of 10,980 US adults and 1,334 Anthropic staff, finds the two groups agree on what AI can do today but split sharply on how fast it will improve.

Key Facts

Finding Figure Report
Junior share of technical staff 21% to 14% since early 2025 Employment
Junior share of technical hires 27% to 9% Employment
Technical teams as share of all staff 61% to 56% in the past year Employment
Interactive agent hours vs. employee hours 6 to 1 by August 2026 Employment
Code commits per engineer About 5x six months earlier Employment
Meeting time, median employee Up 20% from a year ago (16% for technical individual contributors) Employment
Expect AI to make a Nobel-level discovery by 2030 39% of US adults, 91% of staff Expectations
Share of own current work AI could do by 2030 (average) 28% for US workers, 75% for staff Expectations
Expect AI to do at least 80% of own work by 2030 5% of US workers, over half of staff Expectations
A suitable task takes a quarter of the time or less with AI 26% of US adults who use AI at work, 59% of staff Expectations

What the Employment Report Says

The report's authors are Nathan Wilmers, Regina Catipon, Maxim Massenkoff and Peter McCrory. It draws on employment records, agent telemetry, output metrics, employee surveys, Slack and calendar data, hiring records and 17 employee interviews.

Anthropic says engineering and research teams had deemphasized hiring for lower-band technical roles by April 2026. In a survey that month, around a third of machine learning managers and a fifth of software engineering managers said they were hiring less because of AI. Junior job postings fell most sharply after December 2025, and junior employment grew least in jobs where staff use Claude for the most complex tasks.

Agent work hours passed human work hours in March 2026. The 6 to 1 figure counts only interactive sessions directed by employees and assumes 40 human hours per week. Anthropic also tracked 192 engineers who adopted coding agents against 74 who did not, and found code output rose sharply after adoption. It found no difference in how often agent-written code was later deleted or reverted compared with code the same person wrote by hand. The company notes that adoption was not random, so the estimates may be biased.

Human attention is now the limit, the report says. Interactive Claude sessions spend about 80% of their time waiting for a person, and reading all agent output would take more than half the work week even for non-technical staff. "Meta-work", such as writing specs for agents and monitoring pipelines, rose from 10% of the average employee's interactive sessions in fall 2025 to 22% in spring 2026. In hiring assessments, comments on autonomy and judgment rose while comments on domain expertise fell.

Anthropic describes an early warning system that tracks jobs at risk and says it gives priority to moving affected staff into other roles. The report says this does not address the slowdown in junior hiring.

What the Expectations Survey Says

Morning Consult surveyed US adults from August 11 to 23, 2026, with responses weighted to be nationally representative. The staff poll ran August 9 to 12 and is unweighted. In both groups, people who expect AI to do more of their own work forecast higher unemployment in 2030. At the same level of expected automation, both groups forecast about the same unemployment rate.

Anthropic also ran the answers through its "Economic Scenarios for Transformative AI" model. The median US adult's answers imply 5.4% GDP growth in 2030; the median staff member's imply 17.1%. The authors say these are model outputs, not forecasts. Staff answers on hiring and on the share of model-building work done by AI are personal views, not company plans.

Reactions and Impact

TechCrunch highlighted the 20% rise in meeting time and the staff description of managing agents as "babysitting". The Decoder noted that the report itself calls Anthropic a special case: a software company with heavy AI use, fast business growth and voluntary tool adoption, which may skew the numbers. It also cited a paper by Nolan Lovett of the NATO Special Operations University, who argues that cutting entry-level roles now reduces the supply of experienced workers later.

What We Don't Know Yet

  • How much of the junior hiring drop comes from AI rather than Anthropic's rapid growth and changing needs. The report shows the trends together but does not separate them.
  • Whether the patterns hold at companies that are not AI labs.
  • How Anthropic's early warning system works in practice and how many staff it has moved.

Sources