Juniors took 27% of Anthropic's technical hires in early 2025, and now take 9%
Junior roles now make up 9% of Anthropic's technical hires, down from 27% in early 2025. Anthropic published these figures itself. In a research post dated October 16, Nathan Wilmers, Regina Catipon, Maxim Massenkoff and Peter McCrory say that the share of junior-level employees in technical departments fell from 21% to 14% since early 2025. They also say that the share of new technical hires who were juniors fell even faster, from 27% to 9%.
Figure 5 from Anthropic's research post compares the number of junior employees in January 2026 with the number in July 2026. Image: Anthropic
These figures show percentages, not the actual number of people. Anthropic says it is still hiring for technical roles, and the total number of technical employees continues to grow. However, hiring in sales, finance and recruiting has grown faster. Junior job postings are an exception to the wider hiring trend. They have fallen especially sharply since December 2025.
Anthropic says the change began in January 2026
Anthropic says concerns about lower demand for junior technical workers first appeared in key technical teams in January 2026. It says that more capable AI models helped speed up this change in thinking. The company quotes an unnamed employee: "the thing that raised eyebrows was the uplift for performance engineers in particular and it basically overnight changed what we were looking for."
By April, engineering and research teams had put less focus on hiring people for lower-level technical roles. A survey that month also found that most managers were looking for different skills because of AI. Anthropic says that about one-third of machine learning (ML) managers and one-fifth of software engineering (SWE) managers reported hiring fewer people because of AI.
Anthropic says hiring managers are increasingly looking for people who can work independently and make good decisions, rather than focusing mainly on specific technical skills. The company also reports that junior employment is growing more slowly in jobs where people use Claude for the most complex tasks. It presents this as a pattern it observed, not as proof that complex Claude use caused the change.
The artwork Anthropic published with its October 16 research post. Image: Anthropic
Anthropic measures AI agent hours in two different ways
Anthropic compares the time AI agents spend working with the time human employees spend working. However, it describes this comparison in two ways that do not measure exactly the same thing. The company says that by August, interactive AI agents outnumbered employees by six to one. It also says that AI agent work hours were six times human work hours. One statement compares agent hours with the number of people, while the other compares agent hours with human work hours. The post does not explain which meaning is intended.
The human work hours in this calculation are estimates, not hours employees actually recorded. Anthropic says: "Human work hours are calculated as headcount and assumed 40 hours per week."
The large number of AI agents behind this comparison is not new. In September, Anthropic said that about 30,000 agents were working on research and engineering at the same time in August on its most widely used internal platform. In its own account of moving its monorepo away from Git, the company also reported that the number of code changes it merged each day was about 2,500 in May and had risen above 30,000 by September.
Technical employees make up a smaller share of a larger company
Over the past year, technical departments fell from 61% to 56% of Anthropic's total workforce. The company also says it earns more revenue per employee than earlier generations of large technology companies did at a similar stage of growth. At five years old, Anthropic earns between two and four times as much revenue per employee as Meta and Google did when they reached the peak of their own rapid growth.
The mix of workers inside technical teams has also changed. By April 2026, the share of employees who had made no code changes in the previous 30 days had fallen by almost half over the course of a year. The ratio of engineers to product managers is about 17 to 1, and the number of code changes made by managers has doubled. Among employees who had made no code changes at the start of the period, the number of coding-focused AI agent sessions rose from two per month in August 2025 to 90 in May 2026. These sessions were used by employees who did not have any code commits at the start of the period.
Employees are also moving between departments more often. In 2026, employees who had been at the company for at least a year were 50% more likely to change departments in any given month than they were in 2025.
Anthropic says it has responded by creating an early warning system to track jobs that AI could disrupt. As jobs change, the company says it has given priority to moving existing employees into other roles within the company. The system uses measures for each type of role to track demand and productivity, alongside the capabilities and use of AI. Anthropic has not said how many roles the system has flagged or how many employees it has moved to other jobs.
All these findings come from Anthropic's own data about its employees, work output and surveys, as well as interviews with 17 employees. The company presents itself as a "leading indicator of how AI will shift work organization, hiring, and labor demand across the economy". However, Anthropic also sells AI tools, and its research does not measure what is happening in the job market outside the company.