Anthropic's own staff expect AI could do 75% of their work by 2030, US workers 28%, its survey finds

Anthropic asked its own employees and a sample of US adults how much work AI will be able to do by 2030, and the answers are far apart. On average, staff expect AI to be able to do 75% of their current work hours. US workers put it at 28%, according to Anthropic's report, "The AI expectations gap", published Oct. 16.

Who ran the survey, and who answered

The report comes from Anthropic, and one of the two groups surveyed is its own staff. The page says staff "build and use frontier AI every day."

The US sample is 10,980 adults, surveyed by Morning Consult from Aug. 11 to 23 and weighted to be representative of adults. The staff poll was internal, ran Aug. 9 to 12, and drew 1,334 respondents. It is unweighted.

What the gap looks like

On their own jobs, more than half of staff expect AI to be able to do at least 80% of their work hours by then. Anthropic says only "1 in 20" US workers say the same.

On capability, 91% of staff expect AI to be able to make a Nobel-level discovery by that date. Among US adults the share is 39%. The page says both groups largely agree on what AI can do today and differ on how fast it improves.

On time saved, 59% of staff say a task suited to AI takes a quarter of the time or less. Among US adults who use AI at work, it is 26%.

What the authors say about their numbers

The page says the US survey was online with opt-in panelists, so it excludes adults without internet access and may differ from the population in ways weighting cannot correct. Neither group's response rate is given.

Our reading: the staff are self-selected employees of an AI company, and the page gives no response rate, so the size of the gap is partly a question of who answered. The page does not say this explains the gap.

A growth estimate, built by a model

The authors also ran each respondent's answers through their own economic-scenarios model. The median US adult's answers imply 5.4% GDP growth in 2030. The median staff member's imply 17.1%.

Those figures depend on how the authors map survey answers to model inputs. They are model outputs, not forecasts from the respondents. They cover only people who answered every relevant question: 3,259 US adults and 1,093 staff.

Work on the models themselves

Separately, staff were asked about the work of building an early-2026 frontier AI model, not their own jobs. The median estimate of AI's share of that work is 51% today, 80% by the end of 2027 and 95% by the end of the decade.

The page labels these the personal estimates of staff in an internal poll, not a measurement by Anthropic or a company forecast. The same caveat covers staff hiring expectations, which the page says are not company plans.

Jobs and unemployment

Both groups link higher expected AI exposure of their own work to higher unemployment forecasts, the page says. Our reading: that is an association in survey answers, not evidence about what will happen to jobs.