Data visualization charts

Photo by Luke Chesser on Unsplash

The Massachusetts Institute of Technology released research this week that has generated predictable alarm: artificial intelligence can already replace 11.7% of the American labour market, representing approximately $1.2 trillion in wages. According to CNBC’s coverage, the finding comes from the “Iceberg Index,” a simulation tool developed jointly by MIT and Oak Ridge National Laboratory.

One ought to read these studies carefully before succumbing to either panic or dismissal. The researchers are explicit about what their model measures and what it does not. The 11.7% figure represents technical capability and economic feasibility—jobs where AI can perform the tasks at competitive cost. It does not predict that these jobs will disappear on any particular timeline.

The metaphor is apt. According to Fortune’s analysis, current AI adoption concentrates in technology occupations representing just 2.2% of labour market wage value—the visible tip of the iceberg. The remaining 9.5% sits beneath the surface: routine functions in HR, logistics, finance, and office administration where AI capability exists but deployment has not yet occurred.

The simulation tracks 151 million workers across 923 occupations in 3,000 counties, covering more than 32,000 skills. This granularity enables analysis that previous workforce studies lacked. Rather than making sector-level generalisations, the Iceberg Index can identify exposure at the county level, revealing geographic distribution that defies coastal tech hub assumptions.

According to Newsweek’s reporting, Washington, Virginia, and Delaware show highest vulnerability due to concentrated finance and administrative sectors. Rural and inland states show lower exposure but are not immune—the index simulations reveal affected occupations spread across all 50 states.

What makes the MIT research valuable is its policy orientation. The researchers partnered with Tennessee, North Carolina, and Utah to validate the model using actual state labour data. These states have begun building policy scenarios using the platform—testing different interventions before committing resources to workforce training or economic development.

The tool distinguishes between exposure and displacement. Exposure measures technical capability; displacement depends on firm strategies, worker adaptation, and policy choices. A job exposed to AI automation may evolve rather than disappear. Previous automation waves—from manufacturing robots to spreadsheet software—changed work without eliminating it wholesale.

However, the MIT researchers acknowledge that current AI capabilities differ from previous automation technologies. Large language models can generate text, summarise documents, write code, and conduct conversations. These are cognitive tasks previously considered automation-resistant. The historical pattern may not hold.

The $1.2 trillion figure deserves context. American GDP exceeds $28 trillion. The exposed wage value, whilst substantial, represents a manageable fraction of economic output if transition occurs gradually with policy support. The danger lies in rapid, unmanaged displacement that overwhelms retraining capacity and social safety nets.

For individual workers, the Iceberg Index suggests specific attention to skills development in areas where AI capability is advancing. Administrative work, routine analysis, and repetitive communication tasks face highest near-term pressure. Creative work, complex judgment, and physical tasks requiring human presence retain comparative advantage—for now.