The Anthropic Institute has published “The AI expectations gap” (16 October 2026), a technical report by Szymon Sacher, Tess Cotter, Charles I. Jones, Anton Korinek and Peter McCrory. It compares a nationally weighted survey of US adults with an internal poll of Anthropic staff. The headline gap is stark: 39% of US adults and 91% of staff expect AI to be able to make a Nobel-level discovery by 2030. A September re-field of US adults raised that public share to 44%; the report still centres on the August results.
These are Anthropic’s own survey results (fielded with Morning Consult for the public sample). They are not independent forecasts, and several staff answers are explicitly personal views, not company plans.
Method
Per the report’s footnote:
- US adults: Morning Consult surveyed 10,980 US adults between 11 and 23 August 2026, weighted to be representative of the adult population. The survey was administered online to opt-in panelists, so it excludes adults without internet access and may differ from the population in ways demographic weighting cannot fully correct.
- Horizon split: about future outcomes, half the public sample was asked about 2027 and half about 2030; the report uses the 2030 answers unless it says otherwise.
- Staff: an internal poll from 9 to 12 August 2026 with 1,334 respondents from across the company; unweighted; everyone asked about 2030.
- September re-run: Morning Consult fielded the same questions again 17–24 September to 7,613 US adults. Main results “changed little”; the largest change was the Nobel-by-2030 share rising to 44% from 39%. The report says it reports the August results throughout.
Capabilities: agree on today, diverge on 2030
US adults and staff largely agree on what AI can do today, the report says, and rank tasks in nearly the same difficulty order. They disagree on the path ahead. Besides the Nobel gap (39% vs 91%), the authors note that depending on the task, 19–40% of US adults say AI will never be able to do it, versus at most 2% of staff.
Among US adults who use AI at work regularly (about a fifth of respondents), 56% expect AI to do at least seven of eight named tasks by 2030 — higher than all adults (41%) and never-users (28%), but still far from staff (96%).
Own jobs and time savings
On average, staff expect AI to be able to do 75% of their current work by 2030, versus 28% for US workers overall. “Only 1 in 20” US workers say AI could do 80% or more of their work hours by 2030, while over half of Anthropic staff expect at least that level.
Among US adults who use AI at work, 26% say a well-suited task takes a quarter of the time or less with AI, compared with 59% of Anthropic staff.
Unemployment gradient
Respondents were told the unemployment rate is “about 4.2%” today and asked to forecast 2030. US workers who expect AI to do 80% or more of their work expect, on average, an unemployment rate of 12.2% in 2030. Across US workers, “ten more percentage points of work affected is associated with 0.4 points more forecasted unemployment.” The same gradient appears among staff even though their own exposure answers are much higher — the report’s point is that both groups link personal automation to higher unemployment in a similar way.
Adoption and obstacles
Of work tasks AI will be able to do, US adults expect 42% to be done with AI by 2030; staff expect 65%. Asked for up to three obstacles, US adults most often named “Customers preferring to deal with a person”, followed by data privacy/security, reliability, and laws/regulation/liability.
Staff hiring views — personal, not plans
Even while expecting AI to do most of their work, staff still expect Anthropic to keep hiring: 95% expect headcount to grow over the next 12 months; the median expects growth of 66%; 27% expect at least a doubling. The report states plainly: “These are the personal expectations of staff members in an internal poll and do not represent Anthropic’s plans or its view of its headcount.”
Model-implied GDP — Anthropic’s scenario model
Using the Institute’s own framework in Economic Scenarios for Transformative AI, the authors map respondents who answered all key inputs (3,259 US adults and 1,093 staff) into model-implied 2030 GDP growth. The median US adult answers imply 5.4% growth in 2030; the median staff answers imply 17.1%. Those are outputs of Anthropic’s model given survey answers, not official Anthropic GDP forecasts and not measured GDP.
Inside the lab (personal estimates)
Staff median estimate of AI’s share of the work of building an early-2026 frontier model: 51% today, 80% by end-2027, 95% by end-2030 — again “personal estimates of staff members in an internal poll and are not a measurement by Anthropic or a company forecast.”
Why it matters
The report’s useful distinction is not “AI bullish vs bearish” in the abstract but who sits next to the models. Staff and the public agree more on present capability than on 2030; they reason similarly from own-job exposure to unemployment; and staff hiring optimism can coexist with extreme automation expectations. For labour and policy readers, the September 44% Nobel re-run shows public expectations can move quickly even while the August gap remains enormous.
Related TSN context on jobs and Anthropic: Kroger’s planned corporate cuts, Zoho’s “grim” AI-cost outlook, Netflix’s ~6% layoff, Anthropic’s METR evaluator arrangement, and reported Anthropic Q3 revenue.
What’s still unknown
- How much of the staff–public gap is information versus selection into AI jobs
- Whether the September 44% Nobel figure continues to rise in later waves
- External replication with probability samples rather than opt-in panels
What this means
- Anthropic/Morning Consult (Aug): 39% of US adults vs 91% of staff expect Nobel-level AI by 2030; Sept re-run of adults: 44% (report still leads with August).
- Own work by 2030: staff mean 75% vs US workers 28%; ≥80% of hours: over half of staff vs 5% of US workers.
- Time savings (quarter or less): 26% of AI-using US adults vs 59% of staff.
- Staff headcount and frontier-build AI-share answers are personal poll views, not company plans.
- Median model-implied 2030 GDP growth: 5.4% (adults) vs 17.1% (staff) — Anthropic’s scenario model, not a forecast.
Sources
Related stories
- Kroger Plans to Cut About 3,000 Corporate Roles Over Three Years, CEO Memo Shows
- Zoho Skips Its Usual Pay Rises as AI Costs Soar; Founder Sridhar Vembu Says the Outlook for Software Jobs Is “Grim”
- Netflix Lays Off About 6% of Its Staff, Co-CEOs Tell Employees
- Anthropic Says It Has Embedded Six METR Staff With Employee-Like Access and Cannot Veto Their Reports
- Anthropic Told Investors Its Third-Quarter Revenue Reached $16.8 Billion, Reports Say
Follow @tsnmediahub on X for breaking AI news as it happens.
Follow on X
