It's not factory workers or low-wage service staff who face the greatest AI disruption — it's highly educated, well-paid knowledge workers. And women are overrepresented. The Anthropic Economic Index delivers, for the first time, findings grounded in 4 million real AI conversations — not surveys, not models.
The Counterintuitive Truth About AI and Jobs
Who's really at risk when AI reshapes the labor market? The intuitive answer: low-wage workers, the less educated, people doing routine jobs. That answer is wrong — at least when you look at actual usage data.
The Anthropic Economic Index provides, for the first time, an answer grounded not in surveys or theoretical models but in more than 4 million real conversations with Claude.ai. The finding: the most AI-exposed occupations are well-paid, highly educated — and disproportionately female. That's not a minor footnote. It inverts the historical pattern of technological disruption entirely.
The Index rests on two research papers: one analyzing usage patterns across Claude.ai conversations (Handa, Tamkin et al., February 2025), the other translating those patterns into labor market effects (Massenkoff & McCrory, March 2026). The methodological backbone is Clio, Anthropic's privacy-preserving analysis tool that clusters millions of anonymized conversations into thematic groups without exposing individual content. Those clusters were matched against the **O*NET database — a U.S. Department of Labor system covering roughly 20,000 tasks across 800 occupations. The result is a new metric, "Observed Exposure"**: not who hypothetically could use AI, but who actually does — and for what. The full dataset is open-source on Hugging Face.
Klingt interessant?
Numbers That Change the Picture
Start with context: 37.2% of all Claude queries involve Computer & Mathematical tasks — software development, data analysis, algorithms. The largest category by far. But even here, Claude covers only 33% of theoretically possible tasks. The gap between what AI could do and what's actually being used is enormous. We are still early.
Thirty-six percent of all occupations use AI for at least a quarter of their tasks. Only 4% use it for three-quarters or more. Disruption is already real — it's just not yet universal.
Then there's the distinction that most coverage misses: augmentation versus automation. Augmentation means AI works alongside a human; automation means AI handles the task independently. The study finds 57% augmentation against 43% automation. In more than half of all real AI usage, AI isn't replacing people — it's making them more productive. That's the most important number in the study.
The most exposed occupations:
| Rank | Occupation | AI Exposure |
|---|---|---|
| 1 | Computer Programmers | 75% |
| 2 | Customer Service Representatives | 70% |
| 3 | Data Entry Keyers | 67% |
| 4 | Medical Record Specialists | 67% |
| 5 | Market Research Analysts & Marketing Specialists | 65% |
| 6 | Sales Representatives | 63% |
| 7 | Financial and Investment Analysts | 57% |
| 8 | Software Quality Assurance Analysts | 52% |
| 9 | Information Security Analysts | 49% |
| 10 | Computer User Support Specialists | 47% |
75% for Computer Programmers is a striking number. Boris Cherny, the creator of Claude Code, has said publicly he expects the "title of software engineer" to "go away" in 2026. That sounds provocative until you see the data behind it. At the other end of the spectrum: cooks, mechanics, lifeguards, bartenders. Physical labor remains largely untouched. As Massenkoff and McCrory write, many tasks remain beyond AI's reach — "from physical agricultural work like pruning trees and operating farm machinery to legal tasks like representing clients in court."
The Demographic Paradox
This is where the study really upends the historical narrative. The most AI-exposed workers — compared to those with no exposure — are 16 percentage points more likely to be female, 11 points more likely to be white, nearly twice as likely to be Asian. They earn 47% more on average. And they are nearly four times as likely to hold a graduate degree: 17.4% versus 4.5%.
The Industrial Revolution hit low-skill, physical workers, low-wage earners. The AI disruption hits academics, high earners — and, against every prior expectation, women disproportionately. Why women? Because the most exposed occupations — marketing, customer service, documentation, HR, financial analysis — are historically more female-dominated than mechanical engineering or construction. The feminization of white-collar work intersects directly with AI's disruption pattern.
The policy implications reach far beyond tech. Dario Amodei, CEO of Anthropic, was direct: AI could replace "up to half of all entry-level white-collar jobs in the next one to five years." Entry-level jobs. The career ladder that universities implicitly promise graduates. The IMF puts 40% of all jobs globally at risk — 60% in advanced economies. Goldman Sachs estimates the equivalent of 300 million full-time positions affected, while projecting a potential 7% GDP gain if the productivity is captured effectively.
Aggregate employment data looks stable for now: the Yale Budget Lab found no measurable displacement through October 2025. But a Stanford study from November 2025 showed a 16% relative decline in new hires of young graduates in AI-exposed roles. The normalization isn't starting at the top — it's starting at the bottom of the career ladder.
Augmentation, Not Replacement — For Now
The real finding of the Anthropic Economic Index isn't the table of exposed occupations. It's the 57% figure. AI is — as of today — primarily not a replacement engine. It's a leverage tool. More than half of actual AI usage amplifies human work rather than supplanting it.
That will change. At what pace depends on the next generation of models and on the decisions organizations make in the meantime. What tends to hold: complex judgment, genuine client relationships, emotional intelligence, physical execution — areas where AI remains structurally weaker than in text processing. Building toward these strengths creates an advantage that won't be competed away in the next model release.
The uncomfortable takeaway from the Index is this: it's not the bottom of the income ladder that's most exposed first — it's the middle. Highly educated, well-compensated knowledge workers carry significantly higher exposure than anyone previously assumed. That's not a reason for panic. It is a reason to look clearly at where you actually stand. At Nopex, working with organizations through exactly this question, we find the conversation changes the moment exposure stops being abstract: not a threat to manage, but a starting point — which of these tasks can be turned into leverage, and which are worth protecting? The data to answer that already exists. It comes from 4 million real conversations.
