A new simulation from MIT and Oak Ridge National Laboratory is throwing cold water on the idea that AI risk is just a Silicon Valley headline. Their model finds that artificial intelligence already has the technical and economic capacity to perform approximately 11.7% of total U.S. wage value, equivalent to roughly $1.2 trillion worth of work. That shifts the worry away from splashy tech layoffs and straight into routine back-office roles in finance, health care administration, and professional services across the country.
What Project Iceberg Measured
The researchers developed Project Iceberg as a digital twin of the U.S. labor market. The system models 151 million workers, more than 32,000 distinct skills, and thousands of production AI tools to identify where current AI capabilities overlap with the tasks that human workers actually perform. The full methodology and results are detailed in a preprint on arXiv.
Tip Of The Iceberg Vs. What Is Below
In the model, there is a smaller visible "Surface" effect, where AI adoption is concentrated in computing roles worth approximately 2.2% of the wage value, or roughly $211 billion. Below that sits a much larger "Iceberg" exposure at 11.7%, close to $1.2 trillion. MIT stresses that this index reflects technical exposure only, not a countdown clock to layoffs, and is meant as a planning tool for policymakers rather than a prediction of specific cuts.
Which Jobs And Industries Are Most Exposed
Most of that submerged exposure is not in flashy coding jobs but in finance, health care administration, and professional services. Tasks such as document processing, routine financial analysis, and scheduling already align neatly with existing AI tools. National coverage points out that even visible tech roles, such as software engineers, data scientists, and analysts, have started to see some displacement, but a larger share of the wage exposure sits in office and administrative work, according to Fortune.
Not Just Coastal Tech Hubs
Because the Iceberg Index weights skills by wage value, states with large back-office footprints can end up appearing more exposed than traditional tech hubs. The paper notes that states such as South Dakota, North Carolina, and Utah can rank above California or Virginia on the index. That geographic twist is one reason the researchers argue that traditional metrics, such as GDP or unemployment, miss where AI is most likely to reshape local labor markets, and why the index maps exposure all the way down to counties and ZIP codes, as per arXiv.
What This Means For San Francisco
For Bay Area workers, the question is not just whether a headline-grabbing tech company announces a layoff round. It is whether the everyday work that keeps those companies running, from HR and finance to legal operations, billing, and scheduling, becomes cheaper to handle with AI. As reported by KRON4, local leaders and workforce groups are now expected to closely monitor county-level exposure numbers as they inform state planning efforts.
Policy Response And Next Steps
The authors designed Iceberg as a scenario tool, so states and cities can test different interventions before they pour billions into retraining programs and new infrastructure. MIT notes that Tennessee, North Carolina, and Utah have already used the platform to run simulations and set priorities for workforce investments.
The Iceberg Index is not a fate written in code. It measures what AI can do with today’s tools, not what will definitely happen tomorrow. For employers, workers, and policymakers in San Francisco, the study is an early warning to prepare for automation in routine white-collar work, not just the glamorous tech roles that usually make the news.
