#

Key takeaways

• Anthropic studied how AI use changed work inside its own company.

• Junior technical hiring and junior technical share both fell.

• AI agent hours exceeded employee hours by six to one by August 2026.

• Output rose in several areas, but the study warns that output is not the same as value.

• Meeting time increased, which suggests human review became more important.

What Anthropic studied

Anthropic published a study on October 16 about how AI use changed work inside its own company. The authors examined internal employment and productivity patterns. They also included 17 employee interviews. The study is company-specific, so it should not be treated as proof that the same pattern applies everywhere.

The report focuses on how work changed as AI adoption increased. It looks at hiring, department mix, output, and meeting time. It also describes how managers thought about hiring as the company used more AI tools. The authors present the findings as an internal view of one organization, not a market-wide forecast.

Hiring and workforce mix

The study reports a decline in junior-level employees inside technical departments. Their share fell from 21% to 14% since early 2025. Junior technical hires also fell sharply, from 27% to 9%. At the same time, technical departments still grew in absolute headcount.

Technical departments' share of the total workforce fell from 61% to 56%. Anthropic attributes some of this shift to the expansion of business functions. The company also says coding tools may allow nontechnical employees to build software. That is a company-level observation, not a general labor-market claim.

The study also says managers interviewed for the report increasingly emphasized autonomy and judgment when hiring. That suggests the company may value broader decision-making as AI use grows. The report does not say this change was universal across all teams. It only describes the internal pattern Anthropic observed.

AI agent use and productivity

By August 2026, interactive AI agent hours exceeded employee hours by six to one. The study links this rise in agent use with higher output in several areas. It reports more code output per engineer, more sales closed per representative, and more contracts processed after AI adoption.

The authors also caution against reading those output measures too literally. Higher output does not automatically prove higher quality or greater business value. That warning matters because productivity metrics can improve for reasons that do not always show up in final results. The study keeps that distinction explicit.

The report also says human attention has become a constraint. Staff spend time reviewing agent work, which adds to the oversight burden. That means AI can increase throughput while also creating new review tasks. The study presents this as part of the operating reality inside Anthropic.

Meetings and operational pressure

Calendar data in the study show that median meeting time per employee rose 20% over a year. For technical individual contributors, meeting time rose 16%. Those figures suggest that coordination and review demands increased alongside AI use. The report does not say whether the meetings were caused only by AI adoption, but it does connect the trend to the company's changing workflow.

This matters because more AI output can shift work from creation to supervision. The study shows that staff still need time to check agent-generated work. In that sense, AI may change where effort goes rather than simply reducing effort overall. The report frames this as an operational tradeoff inside a highly intensive adopter.

Morocco relevance

The source reports no Morocco-specific facts. A conditional global lesson for readers is that AI adoption can change hiring mix, review workload, and meeting time inside a company. The study also shows that higher output should be checked against quality and business value.

Bottom line

Anthropic's study describes a company that used more AI agents, saw more output in some areas, and hired fewer junior technical workers. It also saw more meeting time and more attention spent on review. The authors are careful to say this is an internal case, not a universal rule. That caution is central to how the findings should be read.