Analysis · 20 October 2026 · 6 min read

How AI changed jobs inside Anthropic

Agents outwork staff six to one and junior technical hiring fell from 27 to 9 percent. What Anthropic's report on its own workforce found, and what it means.

On 16 October 2026 Anthropic published How employment is changing at Anthropic, a report by four researchers at the Anthropic Institute on what happened to work inside the company once AI agents were everywhere. It draws on internal data on headcount, hiring, output, calendars and Slack, plus 17 employee interviews. Anthropic is an extreme case, which is the point: it shows what a fully AI adopted workplace looks like before most companies get there.

Here are the findings that matter for anyone wondering what their own job turns into.

Agents now outwork the humans

By August 2026, the AI agents that individual employees ran worked six hours for every hour their human handlers worked. Counting company wide fleets as well, agent hours passed human hours for the first time in March. Agents appear in nearly half of active public Slack channels, schedule meetings, and brief employees on who they are about to meet.

Output rose, and the researchers checked whether AI caused it

Code commits per engineer ran at five times their pace of six months earlier. Lawyers processed more contracts, accounts payable cleared more invoices, and salespeople closed more deals.

A booming business makes everyone look productive, so the authors compared 192 engineers who adopted coding agents with 74 who had not yet, in the same months. Output rose sharply after adoption. To check for low quality AI code, they compared code written by agents with code the same person wrote directly, in the same month and directory, and found no difference in how often it was later deleted or reverted. They flag the limit themselves: employees choose when to adopt, so some of the effect may be people tooling up for big projects.

Humans became the bottleneck

The striking finding is what the extra output does to people. In interactive sessions, agents spend about 80 percent of their time idle, waiting for a human to respond. Reading everything their agents produce would take even non technical employees more than half their working week.

So the work changed shape. The share of sessions spent on what the report calls meta work, writing instructions for agents, setting one agent to review another, building dashboards to monitor pipelines, rose from 10 percent in autumn 2025 to 22 percent in spring 2026. Standing written instructions for agents went from a few hundred to 8,500 in a year. Asked in an April survey how AI had changed their work, a fifth of employees described their role shifting from doing the work to managing Claudes.

Meetings went up, not down. The median employee spends 20 percent more time in meetings than a year earlier, and even technical individual contributors are up 16 percent. Decisions about systems, priorities and accountability still need people in a room.

Who got hired, and for what

This is where the report touches jobs most directly. The junior share of technical hires fell from 27 to 9 percent, and administrative roles, which usually grow with a company, barely grew. Our post on the pyramid to diamond shift covers what that does to the shape of a company and the career ladder.

What hiring managers look for changed too. Anthropic coded thousands of interview scorecards since 2024:

• Mentions of a candidate's creative use of AI tools tripled for technical roles and quadrupled for non technical ones.
• For technical candidates, the biggest increases were in comments on autonomy and judgment, while discussion of domain expertise shrank.
• For non technical candidates, customer and mission orientation rose and discussion of execution skills fell.

Employees also moved around more. Staff with at least a year at the company were 50 percent more likely to switch departments each month in 2026 than in 2025. One product writing team, whose work was largely handed to agents, moved into writing prompts and evaluations for Claude instead.

The roles whose routine parts the report describes being automated include Software Developers (47) and Lawyers (42), as our index scores them.

What Anthropic says employers should do

The report recommends three things: give employees a financial stake through shared ownership, build an early warning system that tracks which roles AI is changing, and redeploy affected staff internally rather than replacing them. Employee ownership is already widespread at Anthropic, and it says it is building the other two. The authors are candid about the gap: none of this helps the junior candidates who were never hired, and they call for government policy on retraining and income support alongside.

How much to read into it

Anthropic is an AI lab whose employees use the newest models first, own stock, and sell the product they are measuring. Its employees expect AI to be able to do 75 percent of their work by 2030, against 28 percent for other Americans. Its revenue per employee is two to four times what Google and Meta took in at the same age. None of that is typical, and the report is written by the company about itself.

It is still the most detailed evidence published from inside an AI saturated workplace, and its direction matches outside research. It also fits what we see more widely: Anthropic's broader labour research has found no systematic rise in unemployment among exposed workers so far, but a slowdown in hiring at the entry level, which is exactly where this report finds its sharpest change.

Where your job sits

The report's clearest lesson is that the routine, implementation part of a job goes first, and what remains is judgment, coordination and the work of directing AI.

Our index scores 961 occupations from published research on how much of that routine content they carry. The assessment shows how much of your own week looks like it. It takes about two minutes, free, no account.

Figures come from the AI Job Risk Index and were current when this was published. Scores change as new research is ingested, so the index is always the live version. See how scoring works. Informational guidance based on published research, not professional career or financial advice.