The invention of electricity made menial jobs such as lamplighter, elevator operator, and knocker, equivalent to the modern alarm clock, obsolete. The computer made the data entry clerk, keyboard operator, and file clerk redundant.
Anthropic, the artificial intelligence (AI) company that emerged in 2026 as an existential threat to billions of market value, with every ability breathing from its Claude model, is back with a warning about how obsolete AI tools can do all the work. The AI giant, founded by former OpenAI employees who prioritized AI security as progress, has been a thought leader on the risks of AI as progress, and has recently released a study that maps in depth but what tasks AI is actively doing versus what it can only do. The gap between these two numbers is both reassuring and terrifying, depending on your line of work.
In a report titled “The labor market impact of AI: A new measure and preliminary evidence,” authors Maxim Massenkoff and Peter McCrory found that the actual adoption of AI is only a fraction of what AI tools can do.
AI can theoretically cover many jobs in business and finance, management, computer science, math, law, and office management jobs. However, in many cases, actual adoption — measured by researchers using job-related usage data from Anthropic’s AI model Claude — is only a fraction of what’s possible.
Business leaders have for months heeded warnings about AI’s ability to replace white-collar jobs. Anthropic CEO Dario Amodei last year said technology could disrupt half of entry-level white-collar work. The head of Microsoft AI, Mustafa Suleyman, made a similar prediction, predicting that most of the technical work will be replaced within a year to 18 months.
Researchers attribute this lag to existing legal and technical barriers such as model limitations, the need for additional software tools, and the need for people to continue to evaluate the role of AI. But that is temporary, they create.
Who is most at risk?
The research introduces what it calls “exposure exposure” – a new metric that compares AI theory against real-world data, drawn directly from Claude’s experience in technical programs. The discovery jumps off the page: AI is scratching the surface of what it can do with technology. And as it closes that gap, the workforce at risk is older, better educated and better paid.
The workers who would be responsible for that incident are not what most people picture. The group most exposed to AI is 16 percentage points more likely to be female, earns 47% more on average, and is four times more likely to hold a degree compared to the least exposed group. It’s a lawyer, a financial analyst, a software developer, not a warehouse worker. Computer programmers, customer service reps, and data entry keys are the most exposed jobs.
But even those jobs that are most exposed to the capabilities of AI aren’t doing the job of reading right now. The researchers provide an example of what they see as a fully disclosed role that doctors often perform: the authorization of drug refills at pharmacies. AI is capable of automating this task, but they note that they have not seen Claude do it even though it can be completed with a large form of language.
The results are amazing. For computer and math workers, the main types of languages are said to handle 94% of their jobs. But Claude now covers only 33% of these jobs in the recognized use of technology. The same channel exists in Office and Administrative functions – 90% theoretical capacity, a fraction of that in use.
The “red zone,” as defined by the researchers, shows the actual use of AI, which is smaller than the “blue zone” of the possibilities. As efficiency increases and acceptance increases, the researchers write, the red will grow to fill the blue. On the other hand, 30% of workers have zero AI exposure – cooks, mechanics, bartenders, dishes – jobs that require physical presence that no LLM can follow.
Peter Walker, chief information officer at Carta, revealed the blue and red findings in a bar chart. “Universal truth: most radar charts should just be bar charts,” he wrote on X. “Love your stuff, Anthropic!”
The paper cites a scenario that everyone in the knowledge economy should be thinking about: “The great decline of 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 doubling in the high percentage of AI-exposed jobs—from 3% to 6%—would be clearly visible in their system. It hasn’t happened yet, but it absolutely could.
If you think that this is an AI company that is telling its book, this is emerging as a clear opportunity from many perspectives, far from viral doomsday essays like those by Matt Shumer and Citrini Research. Federal Reserve Governor Michael S. Barr put the opportunity among three scenarios he sees for the adoption of AI in a speech last month.
Low pay
The US Bureau of Labor Statistics reported a negative employment report on Friday. Employers shed 92,000 jobs in February and the unemployment rate rose to 4.4%. Some companies have recently announced significant reductions predicted by AI. Jack Dorsey’s Block last month cut nearly half of its workforce, citing AI as the reason. “We’re already realizing that the intelligent tools we’re developing and using, combined with smaller and leaner teams, are enabling a new way of working that’s changing what it means to build and run a company,” Dorsey wrote in an article on X. (Critics including Salesforce CEO Marc Benioff have noted that Block has its own issues,” or that it could be “an AI cause” to do this.)
However, the research finds that for young workers at least, the problem is not dismissal but a decrease in wages in AI-exposed fields, a drop of 14% in the number of job searches in the post-ChatGPT period compared to 2022 in exposed jobs. However, the researchers note that the findings are just as important. And so far there has been no systematic increase in unemployment, according to the research. Citadel Securities, not known for marketing market research, ran a viral doomsday article to find that the hiring of software engineers has increased in recent months.
However, Anthropic researchers suggest that a slight decline may indicate a new reality of work in the age of AI as it reflects some research on the state of the labor market for young workers. A similar study found a 16% drop in jobs exposed to AI among workers aged 22 to 25.
For some young workers, this means ignoring the labor market altogether. “Unpaid young workers may be staying at their current jobs, doing different jobs, or going back to school,” the researchers said.
#Anthropic #mapped #jobs #replace #major #decline #whitecollar #workers #Fortune
