Anthropic's economic team has launched a U.S. economic scenario browser for 2030, accompanied by technical reports from Korinek, Stanford scholar Jones, and others. The model breaks down occupations into O*NET tasks and presents three paths—mild, significant, and extreme—based on capabilities and adoption rates, without asserting that any industry will necessarily disappear by 2030.

In the significant and extreme scenarios, knowledge jobs face large-scale automation, with examples suggesting that coding and call center positions may transition to lower-exposure occupations like electricians and nurses. The report also notes the difficulties of job transition: willingness, skills, and re-employment delays could increase unemployment. In the extreme scenario, cognitive employment without AI is projected to decline by about 21.5%, with a knowledge job unemployment rate of about 17.9% and an overall unemployment rate of about 11.9%; labor's share of GDP is expected to drop from about 60% to about 45%; and non-knowledge job wages could rise by about 33.6% relative to a no-AI path. In the significant scenario, economic growth is expected to roughly double, with knowledge job wages remaining relatively stable. The mild scenario is close to the impact of the internet, with wages remaining stable.

Earlier economic indices ranked computer programmers and customer service representatives among the highest exposure occupations, at approximately 75% and 70%, respectively. CEO Dario Amodei previously stated that entry-level white-collar jobs could disappear by half by 2030, with unemployment rates potentially reaching 10% to 20%. External commentary suggests these figures are more aligned with the model's extreme path. A median expectation from a survey of over 10,000 respondents is close to the significant scenario, with about 10% of respondents anticipating extreme outcomes. The compensation gap for knowledge job income under the extreme path is estimated to be about 9% of GDP.

Mechanically, wages for tasks being automated shift towards capital, primarily computational power; as design and licensing accelerate, demand for on-site construction and nursing increases. Buyers are institutions that need to show policymakers and investors the impact range, while sellers are task models calibrated by Anthropic using their own product usage data. The narrative around funding shifts from "universal wage increases" to stagnation or decline in knowledge jobs, with a scarcity premium for physical and nursing jobs. Beneficiaries include electricians, nurses, and construction workers—occupations less likely to be replaced by models—while coding and telephone customer service jobs, which produce text, face pressure, along with finances that require large-scale transfer payments to offset losses.

Source: Public Information