Anthropic has released an interactive tool that projects how artificial intelligence could reshape the US economy by 2030 — from a near-baseline scenario with modest gains to an extreme one in which cognitive-worker unemployment reaches 17.9 percent. The Econ Scenario Explorer, published by the company's Economics team in September 2026 as version 1.0, lets anyone set assumptions about AI's future capabilities and see the economy those assumptions imply.
The explorer is built on a technical report titled "Economic Scenarios for Transformative AI," published as The Anthropic Institute Working Paper No. 2026-02. Its authors include economists Anton Korinek and Charles I. Jones along with Szymon Sacher, Tess Cotter and Peter McCrory, with Jack Clark providing project direction; the team notes that Claude was used as a research and writing assistant. For more context on this story, see our ongoing AI trends.
How the Model Works
The model represents every job in the economy as a bundle of tasks drawn from the US Department of Labor's O*NET taxonomy. For each task, AI can augment a worker's performance, automate the task outright, leave it unchanged, or create new tasks. Workers are divided into cognitive occupations — management, professional, sales and office roles whose tasks AI can affect — and all other occupations. The cognitive group employed 62.4 percent of workers in the 2025 Current Population Survey.
A scenario, in the paper's framing, is a path for five variables: the share of tasks AI affects, how widely AI use diffuses across those tasks, the productivity gain per AI-performed task, the split between automation and augmentation, and the rate at which new labor tasks are created. When displaced cognitive workers must search for jobs outside their occupation, labor-market frictions can produce sustained unemployment rather than a quick transition.
Anthropic is explicit that the scenarios are not predictions. The company attaches no probabilities to them and describes the framework as a structured way to compare possibilities under different assumptions.
Three Scenarios, Three Very Different Economies
In the modest scenario, 2030 GDP comes in 1.6 percent above its no-AI path — $34.1 trillion at 2025 price levels — with growth of 2.4 percent a year and unemployment of 3.9 percent, barely above the modeled normal of 3.8 percent. The labor share of income slips from 60.0 to 59.4 percent.
The substantial scenario is where the median American's expectations land. GDP runs 8.3 percent above the no-AI path, or $36.3 trillion, and growth over the twelve months to 2030 reaches 5.4 percent a year — a pace the paper sets against the fastest GDP growth of the 1990s dot-com boom, 4.7 percent, recorded in 1999. The costs are uneven: unemployment among cognitive workers rises from 2.9 percent in mid-2026 to 4.5 percent in 2030, cognitive wages end 0.3 percent below their no-AI path, while wages in other occupations run 5.9 percent higher. The labor share falls to 56.1 percent.
The extreme scenario describes an economy transformed. GDP is 32.4 percent above the no-AI path — $44.4 trillion — with growth of 15.4 percent a year, which the explorer says would mean the economy doubling in size every 4.5 years and would likely require recursively self-improving AI systems and rapid adoption for knowledge work. Cognitive-worker unemployment reaches 17.9 percent and economy-wide unemployment 11.9 percent. Cognitive wages fall 11.5 percent below the no-AI path while other wages rise 33.6 percent above it, and labor's share of income drops from 60 to 45.2 percent as capital's share climbs to 54.8 percent.
What 10,980 Americans Expect
Alongside the model, Anthropic surveyed a representative sample of 10,980 US adults through Morning Consult, fielded August 11–23, 2026. Respondents answered questions on when AI will perform eight tasks as well as a skilled professional, how widely it will be used, how large its productivity gains will be, whether it will automate or augment work, and how long a displaced worker would need to find a new occupation.
Running the median respondent's answers through the model produces outcomes close to the substantial scenario: GDP 8.6 percent above the no-AI path by 2030 and unemployment around 4.6 percent. Roughly 10 percent of respondents held views aligned with the extreme scenario. In the survey results, 53 percent said AI can already write routine business emails and documents, 24 percent said it can already build and maintain a working software product, and 39 percent expect a Nobel-level scientific discovery by 2030 — while 40 percent say that will never happen.
The Redistribution Question
The paper's most policy-relevant finding concerns the extreme scenario: a transfer of about 9 percent of GDP — roughly the combined size of Social Security and Medicare — would be needed to hold cognitive workers' income at its no-AI level. The authors note there is no precedent for transfers of that scale in response to technological change.
Anthropic says the model will inform the research it funds through its Economic Futures program on labor-market disruption, as well as the policy ideas it proposes.
Caveats and Outside Review
The company is upfront about the model's limits. Version 1.0 omits policy responses, business cycles, potential aggregate-demand or financial-market disruptions and possible catastrophic risks, and includes no scenarios with advanced robotics — one reason the analysis stops at 2030. Outside economists including Daron Acemoglu, David Autor and David Romer commented on an early draft, though Anthropic says they were not asked to endorse its conclusions.
Open criticisms listed by the company include the model's coarse treatment of individual displaced workers and its omission of the aggregate-demand effects of the data-center buildout. Anthropic says it plans to address many of these criticisms in future versions, and that data observed over the coming years will indicate which scenario the economy is actually in. None of the three, the company argues, can yet be ruled out.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
