Blog
Anthropic survey: its staff expect AI to do 75% of their work by 2030; US workers say 28%
On 16 October 2026 Anthropic published a survey comparing 10,980 US adults with 1,334 of its own staff. Both groups broadly agree on what AI can do today, but 91% of staff expect a Nobel-level AI discovery by 2030, against 39% of US adults. Staff answers are personal views, not company forecasts.
On 16 October 2026 Anthropic published "The AI expectations gap", a report that compares what ordinary Americans and Anthropic's own staff expect AI to do by 2030. The headline result: the two groups mostly agree about what AI can do today, but staff expect much faster progress, wider use and far more of their own jobs to be automated.
The report comes from the Anthropic Institute, the company's research arm.
Who was surveyed
- US adults: the polling firm Morning Consult surveyed 10,980 US adults online between 11 and 23 August 2026. The answers were weighted to be nationally representative.
- Anthropic staff: an internal poll from 9 to 12 August drew 1,334 responses from across the company. These answers are not weighted.
Anthropic notes some limits itself. The public survey used opt-in online panels, so it leaves out people without internet access. The staff answers are personal views, not company plans or forecasts.
The main findings
According to the report:
- Nobel-level science: 91% of staff expect AI to be able to make a Nobel-level discovery by 2030, compared with 39% of US adults. A repeat public survey in September put the public figure at 44%.
- "Never": depending on the task, between 19% and 40% of US adults say AI will never be able to do it. At most 2% of staff say the same.
- Time saved: among US adults who use AI at work, 26% say a suitable task takes a quarter of the time or less with AI. Among staff, the figure is 59%.
- Their own jobs: on average, staff expect AI to be able to do 75% of their current work hours by 2030. US workers put it at 28%. Over half of staff expect AI to do at least 80% of their work, against 5% of US workers.
Jobs and unemployment
Both groups were told US unemployment is about 4.2% and asked to forecast it for 2030. In both groups, people who expect AI to do more of their own work forecast higher unemployment. US workers who expect AI to do 80% or more of their work forecast an average of 12.2%.
Anthropic argues this shows the two groups reason about the labour market in a similar way. What differs is their experience of AI, and so their expectations.
Staff also expect Anthropic to keep hiring: 95% expect headcount to grow over the next year. But they are slightly more pessimistic than the public about how fast a worker who loses a job to AI could find work in a new field. The median staff answer was 12 months; the median public answer was 7 to 12 months.
What it could mean for growth
The researchers ran each person's answers through an economic model from an earlier Anthropic paper. On that model, the median US adult's answers imply GDP growth of 5.4% in 2030, and the median staff member's imply 17.1%. These are model outputs built on expectations, not forecasts of what will happen, and the answers within each group vary widely.
Staff were also asked how much of the work of building a frontier AI model is done by AI. The median answer was 51% today, rising to 80% by the end of 2027 and 95% by the end of 2030.
Why it matters
The report puts numbers on a gap that shapes debates about AI policy and jobs. People who build AI every day expect a much faster and more disruptive future than the public does. Anthropic notes that US adults who already use AI at work hold views closer to its staff, which suggests public expectations may shift as AI use spreads.
Keep in mind that this is an AI company reporting on its own employees' views.