MIT launched Project Iceberg, a population-scale workforce simulator that models 151 million U.S. workers across 923 occupations to measure tasks that current AI can technically perform, according to the project’s report and methodology notes.
The study introduces the Iceberg Index, which finds that 11.7% of total U.S. wage value — about $1.2 trillion — is technically exposed to AI. By comparison, only 2.2% of wage value is currently affected by visible technology adoption. Exposure is concentrated in administrative, financial and professional services but occurs across all regions of the country.
Built on MIT Large Population Models and run on Oak Ridge’s Frontier supercomputer, Project Iceberg is a skills-centered measure of AI’s technical capability, not a forecast of layoffs or timing. A score reflects where AI can overlap with tasks in a given occupation, not whether those tasks will necessarily be automated.
The platform allows states, policymakers and enterprises to simulate policy and training scenarios before committing funding, helping guide strategies for reskilling, redesigning roles and planning AI deployments in back-office workflows.
