New “Iceberg Project” estimates AI could replace 11.7% of U.S. jobs—but reality Is more complex

A new joint study by MIT and Oak Ridge National Laboratory (ORNL)—dubbed the “Iceberg Project”—suggests artificial intelligence could theoretically handle 11.7% of U.S. labor, with the biggest impact in finance, healthcare, and professional services, potentially affecting $1.2 trillion in wages.

Using ORNL’s supercomputer, researchers built a detailed digital model comparing AI capabilities against data from 151 million U.S. workers, analyzing over 32,000 skills across 923 occupations in more than 3,000 counties. The resulting “Iceberg Index” maps where AI could substitute human tasks—even down to individual census tracts—offering granular insights for policymakers.

While the index doesn’t predict exact job losses, it aims to guide local governments on retraining programs and infrastructure investments before disruptions occur. North Carolina Senator DeAndrea Salvador has endorsed the tool for simulating different AI-employment scenarios.

Yet real-world experience tells a nuanced story: many companies that tried replacing staff with AI ended up rehiring original employees, underscoring AI’s current limits. Meanwhile, bipartisan lawmakers are advancing legislation that would require firms to disclose AI-driven job displacements, boosting transparency in the evolving labor market.