The AI Expectations Gap (Anthropic, October 2026): 91% of Anthropic staff expect a Nobel-level AI discovery by 2030

Anthropic's economics team surveyed a nationally representative sample of US adults and its own staff in August 2026 about AI capabilities, adoption, time savings and jobs. The gap is wide on almost every question. 91% of staff expect AI to be capable of a Nobel-level discovery by 2030 compared with 39% of the public. Among US adults who use AI at work, 26% say a task now takes a quarter of the time or less, versus 59% of staff. Staff think AI will be able to do 75% of their own work hours by 2030; US workers put that figure at 28%. One finding is shared: in both groups, people who expect AI to do more of their own job also forecast higher unemployment, so the disagreement is about capability and adoption speed, not about what capability would mean. For founders, the report quantifies how far customers' expectations sit from the frontier. Buyers who believe AI will never do a task need proof, not demos, and the time-savings numbers are a useful benchmark for ROI claims.