MIT puts a number on AI’s impact: 11.7% of US jobs at risk
A new study from the Massachusetts Institute of Technology (MIT), built on a large-scale labour market simulation called the Iceberg Index, claims that existing AI systems are already capable of performing work equal to 11.7% of all jobs in the United States — roughly $1.2 trillion in wages. Behind this dry figure is a detailed “digital twin” of the US workforce that shows where the real shock from AI may come from in the coming years.From hype to numbers: what the 11.7% actually means
The headline figure — 11.7% of the US workforce — refers not to a distant future but to what current AI systems are technically capable of doing today. In other words, if companies fully adopted available AI tools wherever it makes economic sense, tasks equal to almost one in nine jobs could be automated. MIT’s model translates this exposure into money: about $1.2 trillion in annual wages concentrated in sectors such as finance, healthcare, professional services, logistics, administration and back-office support.At the same time, the study does not claim that all of these jobs will vanish overnight. Instead, it describes “technical exposure” — a measure of where AI can already perform core tasks of a role at or above human level. Whether companies move from capability to real-world adoption will depend on costs, regulation, organisational resistance and the availability of reskilling programmes. But the number is large enough to force policymakers, employers and workers to treat AI not as a distant threat, but as an active economic force.
Inside the Iceberg Index: a digital twin of 151 million workers
To get to these estimates, MIT and Oak Ridge National Laboratory developed the Iceberg Index, a labour market model that simulates 151 million US workers as individual agents. Each agent is tagged with their occupation, location, tasks and skills, and is then matched against a catalogue of what existing AI systems can already do. In total, the model spans tens of thousands of skills across hundreds of job categories and thousands of counties, turning a messy labour market into a structured “AI exposure map”.The “iceberg” metaphor reflects how the obvious cases — the visible tip — are only a small part of the story. Tech layoffs, generative AI pilots and coding assistants attract headlines, but they represent just a minor fraction of total wage exposure. The bulk sits underwater, in office roles that never appear on front pages: routine coordination, document handling, reporting, scheduling, standard analysis and other cognitive tasks that AI can quietly absorb without much spectacle.
Visible vs hidden exposure: why office jobs are the quiet epicentre
According to MIT’s analysis, only a small portion of the current AI shock is visible through big tech cuts or automation in obviously digital roles. These account for a few percentage points of total wage exposure. Much larger is the hidden mass of jobs where AI can do 30–70% of the daily workload but has not yet been widely deployed. Examples include HR coordinators, logistics planners, paralegals, junior financial analysts, medical billing staff and administrative assistants.In these jobs, AI tools can already draft documents, summarise long texts, generate standard reports, extract information from forms, route tickets and emails, and support decision-making in routine cases. Humans remain in the loop, but their role shifts from doing the work to supervising and correcting automated output. For employers under cost pressure, the temptation is obvious: keep fewer people, and let AI handle the rest.
Not just Silicon Valley: which regions are most exposed
One of the most uncomfortable findings of the Iceberg Index is that AI risk is not limited to coastal tech hubs. When exposure is measured by skills and tasks rather than by industry labels alone, many smaller states show higher vulnerability than California or New York. Regions whose economies depend on administrative, financial, professional and service roles often have a large share of jobs sitting squarely in AI’s comfort zone.This geographic lens is exactly why the tool is already being used by policymakers in several US states. Instead of assuming that “AI risk” equals “tech layoffs in big cities”, local governments can see which counties are likely to face pressure on wages and employment in the next decade. That, in turn, helps them decide where to invest in training programmes, digital infrastructure and economic diversification before the shock fully arrives.
What this means for workers: automation vs augmentation
For individual workers, the study reinforces a pattern that has become familiar in previous waves of automation: the problem is not AI in general, but the specific mix of tasks in a role. Jobs that are mostly routine, predictable and well-documented are easier to hand over to software, while roles that combine technical knowledge with interpersonal work, creativity, negotiation or on-site physical activity are harder to displace.In practice, this means two possible futures that can coexist within the same occupation. In an “automation-heavy” scenario, AI tools are used primarily to cut headcount, with a smaller number of employees managing higher volumes of work. In an “augmentation-heavy” scenario, AI becomes a force multiplier that allows specialists to focus on complex cases, innovation and strategy, while delegating mechanical tasks to machines. Which path wins will depend on corporate strategy, regulation, worker bargaining power and the availability of alternative employment.
Implications for companies: where the real competitive edge lies
For businesses, the Iceberg Index acts as an uncomfortable mirror. It shows where competitors could already be unlocking efficiency gains with AI — not just in glamorous data science teams, but in back offices, operations and support. Organisations that treat AI purely as a marketing buzzword risk being quietly outcompeted by those who rebuild their workflows around it, redesign roles and invest in targeted reskilling.The study hints at a future where the strategic question for executives is not “Will AI destroy jobs?” but “Which of our tasks are commoditised by AI, and what uniquely human capabilities do we want to double down on?” Companies that can articulate a clear answer — and translate it into hiring, training and product strategy — are more likely to capture the upside of AI instead of being exposed only to its cost-cutting pressure.
Policy stakes: using the map before the earthquake hits
On the policy side, the Iceberg Index provides something governments rarely have in real time: a forward-looking map of where technological disruption is likely to hit, measured not by anecdotes but by granular data on skills and wages. Instead of reacting to layoffs after they happen, lawmakers can identify exposure hotspots years in advance and prepare transition pathways, from regional retraining hubs to incentives for new industries.The MIT team stresses that traditional macro indicators — unemployment rates, GDP per capita, average income — explain only a small fraction of the variation in AI exposure across counties. In other words, a region can look healthy on paper while quietly accumulating a large stock of automatable work. Without new indices tailored to the AI economy, this risk remains invisible until it turns into job losses, wage pressure and political backlash.
Conclusion: a wake-up call, not a final forecast
MIT’s 11.7% estimate is not a prophecy carved in stone. It is a scenario based on what AI can technically do today, assuming aggressive adoption. But as a wake-up call, it is hard to ignore. For workers, it is a signal to invest in skills that complement AI rather than compete with it head-on. For companies, it is a prompt to look beyond small pilots and decide where they genuinely want humans to create value.For policymakers, the Iceberg Index may be one of the first tools that lets them treat AI not as an abstract buzzword, but as a measurable force within the real economy. Whether the eventual number is 8%, 12% or more, the direction is clear: AI is no longer just a story about future robots. It is already a question of who will adapt fastest in a labour market where one in nine jobs is technically up for grabs.
Editorial Team — CoinBotLab