‘An army of Claudes, with every one asking for my attention’: Anthropic calls human attention the bottleneck in the botsitting world to come

Anthropic, an AI company, has released a study revealing the future of work in an AI-dominated world. The study highlights that 80% of the time spent on Claude sessions is due to waiting for human input, making human attention the bottleneck for agent output. Employees are increasingly acting as managers for multiple AI agents, often described as "an army of Claudes, with every one asking for my attention."

AI agents surpassed human working hours for the first time in March, with AI agent hours outnumbering human hours six to one by August. Software engineers' code commits have increased five times over the past six months, and non-technical employees now have personal agents monitoring their email and calendars. Managing these agents has become a job itself, with interactive Claude sessions devoting 22% of employees' time to "meta-work."

Even non-technical employees would spend half their work week reading their agents' output. The median employee now spends 20% more time in meetings than a year ago, as the company's growth and hiring in non-technical roles have increased. As new models are released, employees are tackling harder problems and moving faster. This shift has led to a redefinition of jobs, with employees 50% more likely to move across departments in 2026 compared to 2025.

The hiring process is changing, emphasizing autonomy and judgment over specific technical expertise, and there are fewer entry-level people being hired. Despite these changes, the company has made efforts to identify at-risk jobs and reassign employees. The report also suggests that employees expect AI to complete three-quarters of their work by 2030, a more optimistic view than other U.S. workers.

This transformation could reflect how companies and corporate roles change as AI adoption takes root.