Anthropic, the artificial intelligence (AI) company that emerged in 2026 as an existential threat to billions of market value, with each breathtaking new capability from its Claude model, is back with a warning about just how obsolete AI tools could make whole swathes of work. The AI giant, founded by former OpenAI workers who were obsessed with AI safety just as much as advancement, has been a thought leader on AI risk as much as advancement, and just published a study with the most detailed map yet of which jobs AI is actively performing versus which it merely could perform. The gap between those two numbers is both reassuring and alarming, depending on your line of work.

In a report entitled “Labor market impacts of AI: A new measure and early evidence,” authors Maxim Massenkoff and Peter McCrory found that actual AI adoption is just a fraction of what AI tools are feasibly capable of performing.

AI can theoretically cover most tasks in business and finance, management, computer science, math, legal, and office administration roles. However, in most sectors, actual adoption—which the researchers measured using work-related usage data from Anthropic’s AI model Claude—is just a fraction of what’s theoretically capable. 

Business leaders have for months heeded warnings about AI’s ability to replace white-collar jobs. Anthropic CEO Dario Amodei last year said the technology could disrupt half of entry-level white-collar work. Microsoft’s AI chief, Mustafa Suleyman, made a similar prediction, estimating most professional work will be replaced within a year to 18 months. 

The researchers attribute that lag to existing legal constraints and technical hurdles such as model limitations, the necessity of additional software tools, and the need for humans to still review AI’s work. But that’s just temporary, they project.

Who’s most at risk?

The research introduces what it calls “observed exposure” — a new metric that compares theoretical AI capability against real-world usage data, pulled directly from Claude interactions in professional settings. The finding that jumps off the page: AI is barely scratching the surface of what it’s technically capable of doing. And when it does close that gap, the workers most at risk are older, highly educated and well paid.

The workers who would bear the brunt of that scenario are not who most people picture. The most AI-exposed group is 16 percentage points more likely to be female, earns 47% more on average, and is nearly four times as likely to hold a graduate degree compared to the least exposed group. That’s the lawyer, the financial analyst, the software developer, not the warehouse worker. Computer programmers, customer service reps, and data entry keyers are the most exposed occupations. 

But even those careers most exposed to AI’s capabilities are not quite undergoing a job reckoning just yet. The researchers give the example of what they deem a fully exposed task commonly performed by doctors: the authorization of drug refills to pharmacies. AI can certainly automate this task, but they note they haven’t yet observed Claude performing it even though it can theoretically be completed by a large language model.

The results are striking. For computer and math workers, large language models are theoretically capable of handling 94% of their tasks. Yet Claude currently covers only 33% of those tasks in observed professional use. The same gap exists across Office and Administrative roles—90% theoretical capability, a fraction of that actually in use. 

The “red area,” as the researchers describe it, depicting actual AI usage, is dwarfed by the “blue area” of what’s possible. As capabilities improve and adoption deepens, the researchers write, the red will grow to fill the blue. At the other end, 30% of workers have zero AI exposure — cooks, mechanics, bartenders, dishwashers — jobs requiring physical presence that no LLM can replicate.

Peter Walker, head of insights at Carta, extrapolated the blue and red findings into a bar chart. “A universal truth: most radar charts should just be bar charts,” he wrote on X. “Love your stuff, Anthropic!”

The paper names the scenario everyone in the knowledge economy should be thinking about: a “Great Recession for white-collar workers,” noting that during the 2007–2009 financial crisis, the U.S. unemployment rate doubled from 5% to 10%. The researchers note that a comparable doubling in the top quartile of AI-exposed occupations—from 3% to 6%—would be clearly detectable in their framework. It hasn’t happened yet, but it absolutely could. 

If you think this is an AI company talking their book, this is emerging as a clear possibility from many scenarios, far beyond viral doomsday essays such as that by Matt Shumer and Citrini Research. Federal Reserve Governor Michael S. Barr laid out the possibility among three scenarios he sees for AI adoption in a speech last month. 

The hiring slowdown

The U.S. Bureau of Labor Statistics reported a dismal jobs report Friday. Employers shed 92,000 jobs in February and the unemployment rate ticked up to 4.4%. Some companies have recently announced massive layoffs attributed to AI. Jack Dorsey’s Block last month cut nearly half its workforce, citing AI as a reason. “We’re already seeing that the intelligence tools we’re creating and using, paired with smaller and flatter teams, are enabling a new way of working which fundamentally changes what it means to build and run a company,” Dorsey wrote in a post on X. (Critics including Salesforce CEO Marc Benioff have noted that Block has particular issues of its own and may be “AI washing,” or using this as an excuse to conduct necessary layoffs.)

However, the research finds that for young workers at least, the problem is not layoffs but rather a slowdown in hiring within AI-exposed fields, a 14% drop in the job finding rate in the post-ChatGPT era compared to 2022 in exposed occupations. However, the researchers note those findings are just barely statistically significant. And there has so far been no systematic increase in unemployment, according to the research. Citadel Securities, not known for publishing market research, was moved by a viral doomsday essay to note that hiring for software engineers has actually increased in recent months.

Still, the Anthropic researchers suggest that slight decrease may signal the new reality of employment in the AI age as it echoes other research on job market conditions for young workers. A similar study found a 16% fall in employment in jobs exposed to AI among workers aged 22 to 25.

For some young workers, that means skirting the labor market entirely. “The young workers who are not hired may be remaining at their existing jobs, taking different jobs, or returning to school,” the researchers said.