A new analysis from researchers at MIT and Oak Ridge National Laboratory outlines how current AI systems already match human capabilities across a significant share of the labor market. The Iceberg Index, the study’s central measure, shows that digital AI tools can technically perform tasks linked to about eleven point seven percent (11.7%) of total U.S. wage value. The estimate covers roughly one point two (1.2) trillion dollars in work spread across finance, healthcare, administrative services, and professional roles.
The researchers stress that this figure reflects technical exposure rather than predicted job loss. The index measures where AI systems can perform skills found in existing occupations and maps those capabilities across 151 million workers. It does not attempt to forecast adoption timelines or employment outcomes. Instead, it gives policymakers and businesses a forward-looking view of skill overlap that traditional workforce data cannot capture.
To build the index, the team created a detailed digital representation of the labor market using more than 32,000 skills, 923 occupations, and 3,000 counties. Each worker appears as an agent with a skill profile and geographic location. The same skill taxonomy is applied to more than 13,000 AI-powered tools such as copilots and workflow systems. When combined, these datasets show where human and AI capabilities intersect and how much wage value is tied to tasks that AI systems already demonstrate in practice.
One section of the study focuses on what it calls the Surface Index, a view limited to current visible AI adoption. This portion of the labor market is concentrated in computing and technology roles and represents about two point two percent of wage value, or roughly 211 billion dollars. That cluster captures the most publicized examples of automation in software development and related fields. The broader Iceberg Index expands beyond those areas and reveals that the scale of potential task coverage is much larger and reaches well outside major tech hubs.
The analysis shows that administrative, financial, and professional service jobs account for much of the hidden exposure. These roles rely on cognitive and document-processing tasks that AI tools can already perform. As a result, every state registers measurable exposure even when local economies have small technology sectors. The study points specifically to manufacturing regions where white-collar coordination and support functions show far higher exposure than commonly assumed.
Several states have already integrated the index into early planning efforts. Tennessee, North Carolina, and Utah worked with the research team to test model accuracy and explore how policy choices might influence local outcomes. Officials can use the platform to examine county-level skill patterns and experiment with training programs or workforce investments before allocating significant funds.
The study also compares the index with traditional benchmarks such as GDP, income, and unemployment. These indicators show little alignment with the broader Iceberg Index and explain only a small share of state-to-state variation in exposure. This gap suggests that familiar economic signals may not reflect how AI capabilities intersect with real work, making skill-based measures more useful for anticipating transitions.
The authors note several limitations, including the focus on digital AI tools rather than robotics and the decision to measure technical capability rather than adoption behavior. Even with these boundaries, the index offers one of the clearest views yet of how AI fits into the structure of the modern workforce. The findings point to an economy in which AI reaches far beyond visible technology jobs and into routine tasks across the country, creating a need for workforce strategies that match the scale of the transition.
Notes: This post was drafted with the assistance of AI tools and reviewed, edited, and published by humans.
Read next: EU Member States Agree on Draft Online Child Protection Rules Without Mandatory CSAM Scanning
The researchers stress that this figure reflects technical exposure rather than predicted job loss. The index measures where AI systems can perform skills found in existing occupations and maps those capabilities across 151 million workers. It does not attempt to forecast adoption timelines or employment outcomes. Instead, it gives policymakers and businesses a forward-looking view of skill overlap that traditional workforce data cannot capture.
To build the index, the team created a detailed digital representation of the labor market using more than 32,000 skills, 923 occupations, and 3,000 counties. Each worker appears as an agent with a skill profile and geographic location. The same skill taxonomy is applied to more than 13,000 AI-powered tools such as copilots and workflow systems. When combined, these datasets show where human and AI capabilities intersect and how much wage value is tied to tasks that AI systems already demonstrate in practice.
One section of the study focuses on what it calls the Surface Index, a view limited to current visible AI adoption. This portion of the labor market is concentrated in computing and technology roles and represents about two point two percent of wage value, or roughly 211 billion dollars. That cluster captures the most publicized examples of automation in software development and related fields. The broader Iceberg Index expands beyond those areas and reveals that the scale of potential task coverage is much larger and reaches well outside major tech hubs.
The analysis shows that administrative, financial, and professional service jobs account for much of the hidden exposure. These roles rely on cognitive and document-processing tasks that AI tools can already perform. As a result, every state registers measurable exposure even when local economies have small technology sectors. The study points specifically to manufacturing regions where white-collar coordination and support functions show far higher exposure than commonly assumed.
Several states have already integrated the index into early planning efforts. Tennessee, North Carolina, and Utah worked with the research team to test model accuracy and explore how policy choices might influence local outcomes. Officials can use the platform to examine county-level skill patterns and experiment with training programs or workforce investments before allocating significant funds.
The study also compares the index with traditional benchmarks such as GDP, income, and unemployment. These indicators show little alignment with the broader Iceberg Index and explain only a small share of state-to-state variation in exposure. This gap suggests that familiar economic signals may not reflect how AI capabilities intersect with real work, making skill-based measures more useful for anticipating transitions.
The authors note several limitations, including the focus on digital AI tools rather than robotics and the decision to measure technical capability rather than adoption behavior. Even with these boundaries, the index offers one of the clearest views yet of how AI fits into the structure of the modern workforce. The findings point to an economy in which AI reaches far beyond visible technology jobs and into routine tasks across the country, creating a need for workforce strategies that match the scale of the transition.
Notes: This post was drafted with the assistance of AI tools and reviewed, edited, and published by humans.
Read next: EU Member States Agree on Draft Online Child Protection Rules Without Mandatory CSAM Scanning
