The AI Expectations Gap Between the Public and Anthropic Staff

Summary

Anthropic researchers surveyed 10,980 US adults and 1,334 Anthropic staff in August 2026 to compare expectations about AI’s economic impact. The groups broadly agreed on what AI can do today, but staff expected faster progress: 91% thought AI could make a Nobel-level discovery by 2030, compared with 39% of US adults. Staff also expected AI to be used for 65% of technically suitable work tasks, versus 42% among the public, and reported larger time savings from current AI tools. Anthropic staff estimated that AI could perform 75% of their current work by 2030, compared with 28% among employed US respondents. In both groups, people who expected AI to automate more of their own work also forecast higher unemployment; those expecting AI to perform at least 80% of their work predicted 12.2% unemployment in 2030. A model translating survey responses into economic scenarios implied median 2030 GDP growth of 17.1% for Anthropic staff and 5.4% for US adults. The authors stress that these are expectations, not forecasts or measurements, and note that survey coverage and internal polling have limitations.