Anthropic’s AI-era office: fewer juniors, more meetings as Claude reshapes work

The short answer: more senior, less technical and busier in meetings. Since early 2025, junior employees went from 21% of Anthropic’s technical staff to 14%. The drop in new hiring was steeper still, from 27% of technical hires to 9%, and the report says the sharp fall in junior postings began after December 2025.

Technical teams also make up a smaller slice of the company, 56% of the workforce compared with 61% before. Anthropic gives two reasons. Business teams grew faster as the company scaled, and coding tools now let non-technical employees build software on their own, so not every new tool needs an engineer.

One catch: the report gives shares, not head counts. Because the company tripled in size, a lower junior percentage doesn’t prove fewer junior people work there in absolute numbers. Anthropic hasn’t said either way.

The most striking line in the report is about the machines. Claude agents that employees direct first overtook human active work hours in March 2026. By August, they were logging six times as many hours as the staff themselves.

Output rose along with it. Code commits per engineer climbed about fivefold compared with six months earlier, and Anthropic points to more sales closed per rep and more contracts processed. The company admits that more code isn’t automatically better code or better business results, though it says it hasn’t seen signs that quality is slipping.

Separate figures from Anthropic’s IPO filing show how heavy that use is. Valued at the prices it charges customers, the Claude usage inside the company passed its monthly pay bill, excluding stock awards, in April. By September internal usage had grown more than 250-fold since January, and Claude wrote more than 90% of the code added to its software that month. That’s a retail-price valuation, not cash actually spent.

You might expect AI to free up calendars. At Anthropic the opposite happened. Calendar data shows meeting time per employee up 20% on a year earlier, and up 16% for technical staff.

The report’s explanation is that people have become the bottleneck. Claude produces a flood of text, plans and options, and someone has to read it, sort it and decide what to trust. Employees kept describing in-person coordination as something they can’t do without, because trust and accountability still sit with humans.

Staff described the shift vividly. In an April 2026 survey, 27% said Claude had made them more productive, while 20% said their job had moved from doing the work itself to “managing Claudes.” One engineer likened it to babysitting, with each agent asking for attention at once, and another spoke of constant pressure to move fast and hand everything to Claude.

For people starting out, the warning sign is the entry-level pipeline. A survey of Anthropic managers found about a third of machine learning managers and about a fifth of software engineering managers had cut hiring because of AI, with junior roles hit hardest in both groups. If other fast-growing firms copy that pattern, early-career engineers may face a tougher market.

For everyone else, the lesson is about what the job becomes. Less time drafting from scratch, more time checking, editing and deciding.

There’s an open question the report doesn’t answer: juniors often learn judgment by doing routine work. If AI does that work, companies will need another way to train the next generation of senior staff.

Anthropic is one company, and an unusual one. Its staff build AI for a living, got access to advanced Claude agents before most businesses, and are likely more comfortable handing tasks to software than the average worker. The research came from Anthropic’s own economists, including Nathan Wilmers, using HR records, agent data and employee feedback.

The company also flags its own weak spots. Employees who adopted the tools early may already have been on a rising curve, and output counts don’t measure business impact. Not every big tech firm is going all in on Claude, either. Meta and Microsoft were reported in early October to be steering staff away from Claude toward their own in-house AI.