The people most terrified of losing their jobs to AI may have been looking in the wrong direction all along.
For years, the world assumed automation would come first for factory workers, cashiers, truck drivers and other blue-collar jobs. But a prominent AI company that helps build the very technology reshaping the global economy has uncovered something far more unsettling: the workers now standing closest to the edge are America’s highest-paid, best-educated professionals.
The danger is no longer centered around repetitive manual labor. It is coming for coders, analysts, office professionals and white-collar workers whose careers were once considered among the safest in modern society.
According to the company’s findings, AI is already performing large portions of real-world work inside industries many believed required uniquely human intelligence. In some professions, AI is no longer assisting workers — it is quietly beginning to replace core tasks altogether.
The most shocking part? This is not a prediction about what might happen in the future.
It is already happening now.
A new study by Anthropic, the company behind the Claude AI model, has found that the workers most at risk from AI are not low-wage laborers, but some of the highest-paid and best-educated professionals in the United States (see: Page 8 - 9 [How exposure tracks with projected job growth and worker characteristics])
Unlike previous waves of automation that primarily threatened blue-collar jobs, AI is now targeting “knowledge work” — jobs centered around analysis, writing, coding, and decision-making. In some cases, AI is already handling a significant share of these tasks today, not in some distant future. For example, the study found that nearly three-quarters of a computer programmer’s core job tasks are already being performed by AI systems.
The findings come at a time when fears over AI-driven job losses are growing. While many previous studies focused largely on theoretical predictions about what AI could automate, Anthropic’s research stands out because it draws on real-world usage data from its Claude AI model. As one of the few companies with large-scale access to how people actually use AI in workplaces, Anthropic offers one of the clearest pictures yet of AI’s real impact on jobs.
Which Jobs Are Most Exposed?
According to the study, computer programmers face the highest real-world exposure to AI, with AI already covering around 74.5% of their tasks (refer to Figure 3 of Labor market impacts of AI: A new measure and early evidence). Customer service representatives rank second at 70.1%, driven largely by AI systems handling customer inquiries through company APIs. Other heavily exposed occupations include:
- Data entry keyers — 67.1%
- Medical records specialists — 66.7%
- Financial and investment analysts — 57.2%
What makes these findings especially striking is that these are not traditionally low-skill jobs. Workers in the most AI-exposed occupations earn roughly 47% more than workers in jobs with no AI exposure. They are also far more likely to hold graduate degrees — 17.4% compared to just 4.5% among workers in low-exposure occupations — and are more likely to be female (see Page 9 of “Labor Market Impacts of AI: A New Measure and Early Evidence”; further reference: Page 8 - 9 [How exposure tracks with projected job growth and worker characteristics] For example, people with graduate degrees are 4.5% of the unexposed group, but 17.4% of the most exposed group, an almost fourfold difference.)
The reason is simple: AI is increasingly targeting cognitive and analytical work rather than manual labor.
What This Means for Workers
If you work in programming, finance, customer service, data analysis, or another knowledge-based profession, AI is likely already handling tasks similar to yours. That does not necessarily mean your job will disappear overnight, but it may change how companies hire and structure their workforce.
The study suggests that the first signs of disruption are already emerging among younger workers. Hiring into highly AI-exposed occupations has begun slowing for workers aged 22 to 25, potentially signaling that companies are becoming more cautious about expanding entry-level positions in fields where AI can already perform substantial amounts of work.
Which Jobs Are Least Exposed?
From what we can gather from Anthropic’s research, jobs in these eight categories are least at risk of AI (refer to Page 5-8 of A new measure of occupational exposure; also refer to Page 6 - Figure 2 of Theoretical capability and observed usage by occupational category):
- Construction workers
- Installation and repair technicians
- Production workers
- Transportation workers
- Grounds maintenance workers
- Personal care workers
- Food and serving workers
- Agriculture workers (such as farmers)
Most of these are termed blue-collar roles, and, perhaps surprisingly, several pay extremely well. For example, the average salary of an electrician ranges between $60,000 to pushing $100,000 in the United States, while elevator installers and repairers can make $106,000 without a degree. There are still many tasks that remain beyond AI's reach especially in farming from physical agricultural work like pruning trees and operating farm machinery (refer to Page 7 of “Labor Market Impacts of AI: A New Measure and Early Evidence.).
AI’s Theoretical Power vs. Real-World Usage
Many discussions about AI and employment focus on what AI might eventually be capable of doing. Anthropic researchers Maxim Massenkoff and Peter McCrory instead examined what AI is actually doing in real workplace settings today.
To measure this, they created a system called “observed exposure,” which tracks tasks Claude is actively automating in real-world professional use cases (refer to Page 5 of “Labor Market Impacts of AI: A New Measure and Early Evidence.”). Their methodology also places greater emphasis on situations where AI fully replaces human output rather than simply assisting workers.
The gap between AI’s theoretical capabilities and its current deployment remains large. For instance, the study estimates that AI could theoretically perform up to 94% of tasks in computer and mathematics occupations. Yet in practice, Claude is currently handling only about one-third of those tasks (refer to Page 7 of “Labor Market Impacts of AI: A New Measure and Early Evidence.)
Legal work, another field frequently described as vulnerable to AI, has so far seen relatively limited real-world automation. However, the researchers warn that this may simply reflect the fact that AI has not yet reached its full practical capabilities (refer to Page 7 of “Labor Market Impacts of AI: A New Measure and Early Evidence.).
As AI adoption continues expanding across the economy, they argue, the gap between what AI can do and what it is actually doing is likely to narrow — along with job security in many white-collar professions.
What the Data Says About Job Losses So Far
Despite growing concerns, the study found no major surge in unemployment among highly exposed workers since the launch of ChatGPT in late 2022. This aligns with earlier findings from organizations such as the Federal Reserve Bank of Dallas and the Yale Budget Lab.
However, researchers caution that early warning signs may already be appearing at the entry level of the labor market.
Among workers aged 22 to 25, the rate of starting new jobs in highly AI-exposed occupations has fallen by roughly 14% compared to 2022. Workers over 25 did not show the same pattern (refer to Page 13 of “Labor Market Impacts of AI: A New Measure and Early Evidence.”).
The researchers noted there could be multiple explanations for this trend. Younger workers who are not being hired into these roles may instead be staying in their current jobs, moving into different industries, or returning to school.
For now, occupations such as computer programming, customer service, financial analysis, and medical records management are among the key sectors to watch — not because mass AI-driven layoffs are already happening, but because the data suggests the ground is beginning to shift.
Aaron Colt (aaronafter@hotmail.com) is a content creator under the Newswav Creator programme, where you get to express yourself, be a citizen journalist, and at the same time monetize your content & reach millions of users on Newswav. Log in to creator.newswav.com and become a Newswav Creator now!
The User Content (as defined on Newswav Terms of Use) above including the views expressed and media (pictures, videos, citations etc) were submitted & posted by the author. Newswav is solely an aggregation platform that hosts the User Content. If you have any questions about the content, copyright or other issues of the work, please contact creator@newswav.com.