Anthropic staff expect AI to do more of their work than US workers do

The reported gap

Anthropic's October 16 study reports that staff expected AI to handle, on average, 75% of their current work hours by 2030. Employed US respondents put that share at 28%.

Morning Consult surveyed 10,980 US adults through an online opt-in panel August 11-23, 2026, with demographic weighting. Future questions were split between 2027 and 2030; the work comparison uses employed respondents, not all adults. Anthropic's unweighted internal poll reached 1,334 staff August 9-12. Staff used sliders; public answers used ranges converted to midpoints. Public not-sure answers were excluded by question; staff could skip instead.

Sources: Anthropic — The AI expectations gap; Anthropic — Survey appendix

Our interpretation: make the assumptions discussable

For a manager, the useful starting point is to separate a colleague's expectation from a decision about staffing. A forecast can help identify questions worth investigating. It cannot, by itself, establish which responsibilities a team can safely transfer to a tool. Nor does disagreement tell us which person has understood the future correctly.

Ask people to explain the work behind their estimate. Are they thinking about drafting, checking, coordinating with others or accepting responsibility for a result? Two people may describe the same role while counting different activities. Write those assumptions down before comparing their percentages. Then identify what evidence would persuade each person to revise the estimate.

A planning discussion can distinguish demonstrated capability, permission to use a tool and a workable process for reviewing its results. Progress on one does not settle the others. Keep any local trial tied to an agreed quality standard, and include the effort needed to resolve errors or exceptions. That makes the discussion more useful than asking everyone to agree on a single distant date.

These are editorial suggestions, not interventions tested by the survey. The practical value of an expectations gap is that it exposes assumptions to examine. It does not supply a timetable for replacing people, and it should not become one simply because the respondents work close to the technology.