MIT Report Claims 11.7% of U.S. Labor Can Be Replaced with Existing AI (iceberg.mit.edu)

MIT’s Project Iceberg reports that existing AI could replace about 11.7% of U.S. labor by modeling how large numbers of AI agents coordinate with humans. The research builds an “Agentic US” — a simulated population of 151 million synthetic humans interacting with thousands of AI agents — to test coordination protocols that let human-AI teams tackle complex tasks. Rather than measuring isolated AI abilities, Iceberg studies emergent behavior that arises when many capable agents follow designed protocols, then traces how those capabilities ripple across occupations, industries, and communities. The significance is twofold: technically, it shifts attention from single-model benchmarks to multi-agent coordination protocols as the multiplier that extends AI’s reach; practically, it quantifies exposure across jobs and demographics, giving policymakers and organizations a more granular map of disruption risk. Key details include large-scale agent-based simulation, protocol design as the lever for emergent competence, and occupation-level exposure metrics that reveal uneven impacts. The findings suggest automation risk depends as much on how AIs are orchestrated as on raw capability, underscoring the need for targeted retraining, regulatory planning, and research into safe coordination mechanisms.