MIT Says AI Could Replace Nearly 12% of US Workforce

That represents about $1.2 trillion in wages tied to potentially automatable work across finance, healthcare and professional services.

The estimate comes from MIT’s new labor analysis tool, the Iceberg Index, developed with Oak Ridge National Laboratory. The platform models how 151 million U.S. workers interact with thousands of AI tools to measure technical task exposure rather than forecast whether or when jobs will be eliminated.

The study represents the first attempt to map AI capabilities against the entire U.S. labor market at the county level, CNBC reported Wednesday. Tennessee has already incorporated the findings into its statewide Artificial Intelligence Advisory Council Action Plan, and Utah is preparing a similar report.

Visible AI adoption in computing and technology accounts for only 2.2% of wage value exposure, or roughly $211 billion, the report said. The remaining exposure reflects tasks that AI can perform today, even if no employer has automated them yet.

How the Iceberg Index Measures AI Exposure

The Iceberg Index maps AI system capabilities onto occupational skill requirements. Researchers cataloged more than 13,000 AI tools and aligned them with Bureau of Labor Statistics taxonomies covering 32,000 competencies across 923 occupations in about 3,000 counties.

Each occupation’s skills are weighted by importance, automatability and wage value. A skill is considered automatable when AI tools exist that a language model can use to execute that task. Research co-leader Prasanna Balaprakash of Oak Ridge National Laboratory described the platform as building a “digital twin for the U.S. labor market,” per CNBC’s report.

The index measures exposure, not adoption, workforce displacement or policy impacts. Real-world outcomes depend on employer decisions, regulation, job design and reskilling capacity.

Which Jobs Are Most Exposed

AI deployment is concentrated in technology occupations, employing about 1.9 million workers. Software engineers, data scientists and program managers show the highest degree of overlap with existing AI capabilities. Washington leads the nation with 4.2% exposure in tech roles, followed by Virginia at 3.6% and California at 3%.

The broader 11.7% figure reflects capability in document processing, financial analysis and routine administrative tasks. Exposure extends to routine functions in human resources, logistics, finance and office administration, CNBC reported.

The study distinguishes between task replacement and task augmentation. Some AI systems can fully automate specific functions, while others support workers by handling repetitive or low-judgment components. Many exposed jobs may shift toward hybrid models rather than vanish.

MIT economist David Autor told an audience at the MIT AI Conference in March that AI will likely augment human expertise more frequently than replace it.

How Exposure Varies Across Regions and Industries

Exposure is not concentrated solely in coastal technology hubs. South Dakota, North Carolina and Utah show higher index values than California or Virginia when administrative and financial sectors are included.

Industrial states also show high cognitive-task exposure. Tennessee registers 11.6% while Ohio reaches 11.8%, driven by administrative coordination and professional services embedded within manufacturing supply chains. The study used the Herfindahl-Hirschman Index to examine whether exposure clusters in specific industries or spreads broadly. Northeastern states tend to show concentrated exposure led by finance and technology, while Manufacturing Belt states display more distributed patterns across logistics, production, administration and services.

MIT researchers also developed an interactive simulation environment that allows states to model policy scenarios, including training investments and workforce funding before implementation.

The PYMNTS Intelligence report “No Roadmap, No Problem: How Enterprises Are Reinventing the AI Workforce” found that readiness levels for AI adoption inside firms vary. Overall, 60% of chief financial officers said their companies are at least somewhat prepared to manage AI-driven change, including 12% who said they are very prepared. Nearly all the rest, about 38%, were neutral, signaling uncertainty about their ability to adapt.

For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.