Two groups looking at the same future
Anthropic has compared expectations about AI among its staff and a nationally representative sample of US adults. The surveys, conducted in August 2026, show that both groups anticipate substantial progress, but staff expect considerably faster changes in capabilities and work.
On average, Anthropic staff expected AI to be able to perform 75% of their current work by 2030, compared with 28% among US workers overall. The contrast describes what respondents believe will become possible. It does not measure how much work is already automated or confirm what employers will choose to do.
Experience and expectations differ
The gap also appears in expectations about scientific discovery. The report says 91% of staff expected a Nobel-level AI discovery by 2030, compared with 39% of the public. Within current workplace AI use, staff were more likely to report completing tasks in a quarter of the time or less, including checking and fixing the output.
These are different populations with different exposure to AI. The public sample contains 10,980 respondents and is weighted to be nationally representative; staff responses are unweighted. The comparison therefore should not be treated as an experiment showing that a particular amount of AI use causes a particular belief.
What the gap can tell a workplace
The report also records barriers including reliability, privacy and customer preferences. Expectations about technical capability do not settle those practical adoption questions. Staff views about future headcount likewise represent personal expectations in an internal poll, rather than an announced company hiring plan.
Franklin's reading is that the study can help teams identify assumptions they have not discussed. Ask separately what a system may become capable of, what employees would trust it to do, and what customers will accept. Keeping those questions distinct makes the survey more useful than turning either group's answers into a fixed forecast.