How employment is changing at Anthropic

What happened

Anthropic looked at its own workplace. It found that AI agents now do a huge share of the work. These agents are computer programs that can act on their own. They write code, fix bugs, answer chats, and book meetings. Employees now produce more code and close more sales than before. At the same time, the company is hiring fewer junior technical workers compared with its size.

Key numbers and facts

• By August, AI agents worked about six hours for every one hour from employees, in interactive work.
• Anthropic brings in 2 to 4 times more revenue per employee than Meta and Google did at the same age, at their fastest growth.
• The study used company data and interviews with 17 employees.
• Even non-technical staff said that reading all of their Claude agents' output would take about half their work week.

Background: why this matters

Claude is Anthropic's AI model. Many workers now use it to write code and check work. The study says that as AI got better in January 2026, teams that used it most saw that tasks like writing code and finding bugs could be given to AI. One manager said the change happened almost overnight.

Why it matters

Anthropic may be an early sign of how work will change in other companies. The main limit is now human attention. People must check the output of many AI agents, so they spend time watching and organizing the agents. Hiring managers also care more about whether people can work on their own and use good judgment, not only about specific technical skills. This raises hard questions about what companies should promise workers as AI changes their jobs.

What to watch

• Whether other companies also hire fewer junior technical workers.
• How Anthropic's early warning system and internal job moves work for employees whose jobs change.
• Whether human attention stays the main limit as AI agents grow more powerful.