Inside the MIT Data: The 11.7% of the US Workforce Exposed to AI Automation

Inside the Frontier supercomputer at Oak Ridge National Laboratory, a digital replica of 151 million American workers is currently interacting with more than 13,000 artificial intelligence tools. This massive simulation, known as Project Iceberg, was built to solve a fundamental measurement problem. Traditional economic metrics like GDP and unemployment track disruption only after it has happened. To get ahead of the AI transition, researchers needed a way to measure technical capability before adoption crystallizes.[1][2]

Developed by researchers at the Massachusetts Institute of Technology and Oak Ridge, the initiative uses Large Population Models to map the exact overlap between human skills and machine capabilities. The system breaks down the U.S. labor market into 923 distinct occupations and over 32,000 specific skills, testing each one against what current AI models can reliably execute.[1][2]

The resulting data reveals a massive measurement gap in how the public perceives automation risk. The highly visible disruption in software engineering and coastal tech hubs—what researchers call the "Surface Index"—represents just 2.2% of total wage value, or roughly $211 billion.[1]

Beneath that surface lies a much larger hidden mass of cognitive automation. The "Iceberg Index" shows that AI systems can already perform tasks equivalent to 11.7% of the U.S. labor market.[1][3]

This translates to approximately $1.2 trillion in wages tied to work that machines can technically execute today. Rather than coding or data science, this exposure is heavily concentrated in routine administrative services, financial reporting, healthcare administration, and logistics.[1]

Because these cognitive tasks serve as the connective tissue of the broader economy, the geographic risk profile looks entirely different from past technological waves. The exposure is not confined to Silicon Valley or New York; it is distributed evenly across all 50 states, including industrial centers in Ohio, Michigan, and Tennessee.[1][3]

However, the researchers emphasize a critical distinction: technical exposure is not the same as job displacement. The index measures the overlap in capabilities, not a prediction of mass layoffs.[1]

When an AI tool proves capable of performing 30% of a financial analyst's tasks, the ultimate outcome depends on corporate strategy and workflow design. Companies may choose to reduce their headcount, or they may use the technology to augment the worker, absorbing the routine cognitive load so the employee can increase their total output.[3]

This forward-looking data is already shifting how governments prepare for the economic transition. Because the simulation can drill down to the county and ZIP code level, policymakers can identify exact exposure hotspots before local economic pain materializes.[1][2]

Several states are already using the platform as a sandbox for workforce interventions. Tennessee formally adopted the Iceberg Index in its 2026 AI Action Plan to monitor labor market exposure and test the potential impact of reskilling programs before committing public funds.[1]

By mapping the fault lines of the AI economy today, the data provides a rare strategic window. Instead of reacting to headlines about job losses, institutions finally have the granular intelligence needed to upgrade the workforce before the disruption fully scales.[2][3]

The ultimate value of the Iceberg Index lies in its transparency. By quantifying exactly which skills are vulnerable and which remain uniquely human, it transforms abstract anxiety about artificial intelligence into a concrete, navigable map for the future of work.[3]