The invention of electricity made the light bulb, the elevator operator, and the manual, the human equivalent of the modern alarm clock, as trivial tasks. Computers have made the data entry clerk, switchboard operator, and file clerk obsolete.
Anthropic, the artificial intelligence (AI) company that emerged as an existential threat to a market worth billions in 2026, with every exciting new capability from its cloud model, is back with a warning about how obsolete AI tools could make a whole lot of work. The AI giant, founded by former OpenAI employees who are as interested in AI safety as progress, has been a thought leader on AI risk as progress, and has just published a study with a more detailed map of what AI is actively doing than what it can do alone. Depending on your line of work the gap between these two numbers is both safe and dangerous.
In a report titled “The Labor Market Impact of AI: A New Measure and Preliminary Evidence,” authors Maxim Mesenkoff and Peter McCrory find that the true adoption of AI is only a fraction of what AI tools are potentially capable of doing.
AI could theoretically cover many jobs in business and finance, management, computer science, math, legal, and office administration roles. However, in many sectors, actual adoption—which the researchers measured using work-related usage data from Anthropic’s AI Model Cloud—is only a fraction of what is theoretically capable.
Business leaders have heeded warnings about AI’s ability to replace white-collar jobs for months. Anthropic CEO Dario Amudi said last year that technology could disrupt half of the entry-level white-collar jobs. Microsoft’s head of AI, Mustafa Suleiman, predicted the same, estimating that most professional work will be replaced within a year to 18 months.
The researchers attribute the delay to existing legal hurdles and technical hurdles such as model limitations, the need for additional software tools, and the need for humans to still review AI work. But this is only temporary, they project.
The study introduces what it calls “demonstrated exposure”—a new metric that compares theoretical AI capabilities with real-world usage data, extracted directly from cloud interactions in professional settings. The finding that jumps off the page: AI barely scratches the surface of what it’s technically capable of doing. And when that gap closes, the workers most at risk are older, more educated and better paid.
The workers who will bear the brunt of this scenario are not the ones most people picture. The most AI-exposed group is 16 percentage points more likely to be female, earn 47% more on average, and have nearly four times as many graduate degrees as the least exposed group. It’s a lawyer, a financial analyst, a software developer, not a warehouse worker. Computer programmers, customer service representatives, and data entry clerks are the most exposed occupations.
But even the careers most exposed to AI capabilities are not yet in the job equation. The researchers cite an example of what they consider a fully disclosed duty typically performed by physicians: authorizing pharmacies to refill medications. AI can certainly automate this task, but they note that they have yet to see the cloud do it while it could theoretically be accomplished by a large language model.
The results are remarkable. For computer and math workers, large language modules are theoretically able to handle 94% of their tasks. Yet the cloud currently covers only 33% of these functions in observed professional use. A similar gap exists across office and administrative roles—90% theoretical capacity, only a fraction of which is actually used.
The “red field,” as the researchers describe it, depicts the actual use of AI, dwarfed by the “blue field” of what is 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 — cooks, mechanics, bartenders, dishwashers — jobs that require a physical presence that no LLM can replicate.
Peter Walker, Head of Insights at Carta, added the blue and red findings to the bar chart. “A universal truth: most radar charts should just be bar charts,” he wrote in X. “Love your stuff, Anthropic!”
The article names a scenario that everyone in the knowledge economy should be thinking about: “The Great Recession for White Collar Workers” noting that during the 2007-2009 financial crisis, the US 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 it’s an AI company talking about their book, it emerges as the clearest possibility out of many scenarios, far from viral doomsday articles like those by Matt Schumer and Saturni Research. Federal Reserve Governor Michael S. Barr outlined the possibility in a speech last month among three scenarios he sees for AI adoption.
The U.S. Bureau of Labor Statistics released a jobless job report on Friday. Employers lost 92,000 jobs in February and the unemployment rate rose to 4.4%. Some companies have recently announced a wide range of jobs attributed to AI. Jack Dorsey’s block cut half of its workforce last month, citing AI as the reason. “We’re already seeing that the intelligence tools we’re building and using, paired with smaller and smarter teams, are creating a new way of working that fundamentally changes what it means to build and run a company,” Dorsey wrote in a post on X . Dismissal.)
However, the study found that, at least for young workers, the problem is not a slowdown in hiring in AI-exposed fields, with a 14% decrease in employment rates in the post-ChatGPT period compared to 2022 in exposed occupations. However, the researchers note that these findings are only statistically significant. And according to the research, there has been no systematic increase in unemployment so far. Citadel Securities, which is not known for publishing market research, was moved by the Viral Resurrection article to note that hiring for software engineers has actually increased in recent months.
Still, Anthropic researchers suggest that the slight decline may signal a new reality of work in the age of AI as it calls for more research into labor market conditions for young workers. A similar study found a 16% drop in jobs among 22- to 25-year-old workers exposed to AI.
For some young workers, this means that the job market is completely gone. “Young workers who are not employed may stay in their current jobs, take on different jobs, or return to school,” the researchers said.
This story was originally featured on Fortune.com
