News
Anthropic reports agents surpassed employees in hours worked by March 2026 and were six times more numerous by August, alongside a wider expectations gap.
Anthropic says AI agents have moved from a supporting tool to a larger share of its own work than employees, while hiring has shifted away from junior roles. The company also says its staff expect AI to reshape work more quickly than the broader public, underscoring how direct exposure to AI systems is affecting views of the labor market.
Employment shifts inside Anthropic
In a report published Oct. 16, 2026, Anthropic said that over the past year it has hired for a smaller share of junior positions and that a smaller share of employees now work in technical roles. It said the rate of hours worked by AI agents rose sharply, with agents working more hours than employees by March 2026. By August, Anthropic said there were six times more agents than employees at the company.
The company said those changes coincided with higher productivity, including more code produced per person, though it noted that code output is a “notoriously imperfect measure.” Anthropic also said it has closed more sales per sales representative and processed more contracts. In internal analysis tracking the same employee over time, it said output rises as AI tool adoption increases. The company added that the growing acceleration has made human attention the bottleneck, pushing employees to spend more time building systems to monitor agent work.
How staff expectations differ from the public
Anthropic said a second report compared employee views with an August survey of almost 11,000 U.S. adults conducted by Morning Consult and weighted to represent the adult population. The company said the gap is not mainly about different theories of how AI affects jobs, but about different firsthand experiences of AI capabilities.
According to Anthropic, both groups expect AI capabilities to improve, but staff expect faster progress. The company said 39% of U.S. adults and 91% of staff expect AI will be able to make a Nobel-level discovery by 2030. It also said more than half of Anthropic staff expect AI to be able to do at least 80% of their current work by 2030, compared with 5% of Americans. Anthropic said there is still overlap between the groups: two in 10 staff had expectations similar to those of a typical U.S. adult.
Anthropic said the public and its employees agree on one point: within each group, respondents who expect AI to do more of their own work also anticipate larger increases in unemployment. On that basis, the company said both groups appear to share a similar economic model of labor-market effects, even if their personal experience with AI leads them to different expectations for the future of work.
Why it matters
Anthropic said its own workplace may serve as a leading indicator for how AI could affect work organization, hiring and labor demand more broadly, while also cautioning that an AI lab is not the same as a traditional business. The reports offer a company-level view of how adoption at the frontier may be changing the mix of jobs, the pace of output and employees’ expectations about what AI can do next.
Be the first to comment