Anthropic published two reports on AI’s impact on work. One covers how AI changed hiring and productivity inside Anthropic. The other compares public and staff expectations for AI’s future economic impact. Anthropic now hires fewer junior and technical roles. Agent hours surpassed employee hours by March 2026. By August 2026, agents outnumbered employees six to one. Productivity rose in code output, sales, and contract processing. Human attention is now the main bottleneck. Anthropic staff expect faster AI progress than the public. 91% of staff expect Nobel-level AI by 2030 versus 39% of US adults. 5% of US adults expect AI to do 80% of their work by 2030. Both groups share similar economic models of AI’s labor impact. The findings may preview broader workplace shifts.
What changed
Anthropic previously hired more junior and technical roles.
What it unlocks
Employees spend more time monitoring AI agent work due to attention becoming the bottleneck.
- 6x more agents than employees by August 2026
- 39% of US adults expect Nobel-level AI by 2030
- 91% of Anthropic staff expect Nobel-level AI by 2030
- 5% of US adults expect AI to do 80% of their work by 2030
Sources