The jobs most at risk from AI automation aren't necessarily the ones dominating the headlines. Yes, programmers are being laid off. Yes, entry-level tech roles are drying up. But according to a new study out of MIT, the anxiety consuming the tech industry accounts for roughly one-fifth of the actual economic exposure. The other four-fifths — about $1.2 trillion in annual wage value — is sitting in industries that haven't generated a single worried headline. And the workers in those industries have no idea what's coming.
Which Jobs Are Really Most at Risk From AI?
Here's the uncomfortable answer: it's not who you think. The workers facing the deepest AI exposure tend to earn 47% more than the least-exposed group, are nearly four times as likely to hold a graduate degree, and are 16 percentage points more likely to be female. These are people whose working day is built almost entirely around reading, writing, analyzing, and summarizing information — financial analysts, HR coordinators, legal secretaries, insurance claims processors, and a long list of other white-collar professionals who did exactly what society told them to do and did it well.
01:45
How AI replacing tasks inside jobs differs from replacing the jobs themselves
Watch at 01:45 →
The reason this catches people off guard is simple: we've been measuring the wrong thing. Most headlines treat AI as something that replaces jobs. But that's not actually how it works. More often, AI replaces the tasks inside jobs. A lawyer doesn't disappear overnight — but the hours they spend reviewing routine contracts quietly shrink. A journalist keeps writing, but the time spent pulling background research starts to compress. That distinction sounds subtle, but it changes everything about where the real risk lives.
What Is the MIT Iceberg Index and How Does It Work?
The Iceberg Index is MIT's attempt to build a measurement tool actually designed for the AI era — because the ones we already have aren't up to the job. GDP, unemployment figures, and wage data are all built to count jobs and people. They were never designed to look inside a job and ask which parts of it AI can already technically perform.
To build the index, the researchers started by creating a digital model of 151 million American workers across 923 occupations and 3,000 counties. They mapped the skills each occupation requires using O*NET, a US Department of Labor database that breaks hundreds of jobs down into their actual component tasks — things like analyzing data, critical thinking, document processing, and coordinating with others. Each skill comes with an importance rating and a difficulty level drawn from surveys of real workers.
04:20
The O*NET skill taxonomy used to map both human workers and AI tools
Watch at 04:20 →
Then they did the same thing for AI, cataloging more than 13,000 real production-ready AI tools currently being deployed inside companies — from coding assistants and document processors to financial analysis software and workflow automation platforms. Crucially, every one of those tools was mapped through the same O*NET skill taxonomy as the human workers.
The result: for the first time, you can make a genuine apples-to-apples comparison between what human workers actually do and what AI systems are technically capable of doing right now. The index produces a single number for each occupation — a percentage measuring how much of the wage value inside that job AI can technically perform. Weighting by wage value rather than task count is the key design choice. Automating 60% of someone's time doesn't mean automating 60% of their value, and the index is built to reflect that difference.
How Much of the Economy Can AI Actually Replace?
This is where the iceberg metaphor earns its name. When you measure the work AI can technically perform across just the tech sector, it accounts for about 2.2% of total US labor market wage value — roughly $211 billion. That's the visible tip sitting above the waterline, and it's what almost every policy paper and anxious op-ed has been focused on.
07:10
The jump from 2.2% tech exposure to 11.7% economy-wide — the iceberg visualized
Watch at 07:10 →
When you apply the same methodology to the whole economy, the number jumps to 11.7% — roughly $1.2 trillion, five times larger. The same capabilities that make a coding assistant useful to a software engineer — document processing, routine analysis, data handling — overlap heavily with the work done by financial analysts, HR coordinators, legal secretaries, and hundreds of other professional roles that would never appear in a headline about AI layoffs.
The gap between what AI can technically do and what it's actually doing in practice is still large. For computer and math workers, AI is theoretically capable of handling around 94% of their tasks, but in observed professional use it's currently doing about 33%. Similar patterns show up in legal work and architecture and engineering. The technical capability is there — it's being held back by regulation, integration challenges, and the simple fact that most organizations still require a human to check the AI's work. These friction points tend to resolve as technology matures.
The leading edge of this exposure is already showing up in hiring data. Entry-level employment in AI-exposed occupations has dropped 14% compared to the pre-ChatGPT era. Job postings fall before employment does, and entry-level postings across the US have dropped 35% since January 2023.
10:35
State-by-state Iceberg Index scores vs conventional economic vulnerability measures
Watch at 10:35 →
Which US States Face the Biggest AI Job Risk?
If you had to guess which states are sitting on the biggest fault lines, you'd probably say California, Washington, and New York. You'd be wrong. According to the Iceberg Index, South Dakota, North Carolina, and Utah show higher exposure values than California or Virginia.
The reason is economic concentration. California's workforce is diversified enough that AI exposure spreads thin. But states built around administrative and financial services carry concentrated vulnerability. Tennessee makes this point most starkly: its tech sector exposure is just 1.3% — nothing that would trigger alarm in any standard workforce planning model — but its Iceberg Index sits at 11.6%. The white-collar workforce keeping Tennessee's factories running is ten times more exposed than the tech sector everyone has been watching. Ohio and Michigan follow the same pattern. These states have spent years preparing for robots to take over the factory floor. The white-collar disruption is arriving first.
This matters enormously for policy, because the standard economic metrics — GDP, per capita income, unemployment — explain less than 5% of the variation in Iceberg Index scores across states. In some cases the relationship even flips. States that look safest by conventional measures aren't necessarily the least exposed. The billions currently being spent on workforce preparation may be systematically aimed at the wrong places.
Why White-Collar Workers Are More Exposed Than Factory Workers
The conventional story about AI and automation has always featured factory workers and truck drivers as the primary victims. The Iceberg Index turns that story upside down. Cognitive and administrative work — the kind done in offices, not on floors — is far more exposed to current AI capabilities than most physical labor.
Physical work requires embodied presence, dexterity, and real-world judgment that no current AI system can replicate at scale. Writing a contract, processing an insurance claim, or summarizing a legal brief, on the other hand, maps cleanly onto what large language models are already very good at. The jobs that look safe — the ones that generated the most confident reassurances from economists over the past decade — turn out to be sitting directly on the fault line.
Are Jobs AI Can't Replace Actually Safe?
About 30% of the workforce has essentially zero AI exposure — cooks, mechanics, nurses, plumbers, bartenders, childcare workers. People doing physical, relational, hands-on work that no language model can replicate. Surely those workers are fine?
Not exactly. There's an economic pattern that has been quietly working against them for decades, long before anyone heard of a large language model.
How AI Makes Baumol's Cost Disease Much Worse
In 1965, Princeton economist William Baumol noticed something strange about the performing arts. A string quartet performing Beethoven in the 19th century required four musicians and about 25 minutes. A century later, it still required four musicians and about 25 minutes. Nothing had gotten more efficient. And yet the cost of putting on that concert had risen dramatically — dragged upward by wages rising everywhere else as manufacturing, agriculture, and industry became more and more productive.
Economists call this Baumol's cost disease. The things that got more productive — electronics, computers — got cheaper. The things that couldn't get more productive — childcare, education, healthcare, skilled trades — kept getting more expensive. You can't make a nurse see patients faster. You can't make a plumber fix a pipe remotely. The only way to cover their rising wage bill is to raise prices.
AI is about to accelerate this dramatically. If cognitive and administrative work becomes dramatically more productive — a financial analyst compressing a day's work into an hour, a software engineer doing the work of three — then the relative cost of work AI can't touch keeps rising. Healthcare, education, elder care, and skilled trades are things people can't simply stop using when prices go up. And most of them are either funded or subsidized by governments already stretched thin.
Workers who are safe from AI disruption may find themselves in industries that governments will increasingly struggle to afford. Being on the wrong side of the iceberg might turn out to be safer than being on the right side — but it won't feel that way when the bills arrive.
What Happens If We Keep Using the Wrong Map?
The Iceberg Index is ultimately an argument that we're navigating a new city with a 30-year-old map. The streets look familiar, but nothing is quite where you'd expect. Governments and companies are making billion-dollar workforce decisions using tools that can't see 95% of the problem they're trying to measure.
Transitions like this have historically created as many new roles as they disrupted. But the ones that went best were the ones where people could see clearly enough to prepare. Right now, the map most people are using was drawn for a different economy. The Iceberg Index is an attempt to draw a better one. Whether anyone actually uses it is, as always, another question entirely.
