Key insights

• Anthropic quantifying its own product's labor impact is unusual. Most AI companies avoid publishing displacement estimates for the technology they sell.
• The gap between what AI can theoretically do and what companies actually use it for suggests disruption is gradual, not sudden. But junior roles face pressure first.
• 77% of Gen Z prioritizing automation-proof careers signals a generational shift toward skilled trades that could reshape workforce pipelines.

In Brief

Anthropic, the AI safety company behind Claude, published a research paper titled "Labor market impacts of AI" on March 5, 2026. The study, authored by Maxim Massenkoff and Peter McCrory, ranks occupations by how much of their work AI can theoretically perform. CBS News reported that white-collar jobs are the most vulnerable, while hands-on physical jobs remain largely unaffected. The findings arrive as 77% of Gen Z say it is important that their future job is difficult to automate.

---

Most and least exposed jobs

The study looked at AI's theoretical capabilities alongside its current real-world applications. Researchers examined each occupation as a collection of tasks and skills, then measured how capable AI is at performing them. The result is an "AI exposure" score for each job, indicating how much of the work AI could handle.

The most exposed occupations are exactly what you might expect. Computer programmer, data entry specialist, and customer service representative top the list. For computer programmers specifically, the paper estimates that AI can cover roughly 75% of their tasks.

On the other end of the spectrum, jobs requiring physical presence and hands-on skills rank as least exposed. Lifeguard, cook, and motorcycle mechanic were among the safest. The pattern is clear: if the job keeps you away from a keyboard and requires you to use your hands, AI is not coming for it anytime soon.

---

Actual job losses: limited evidence, early warnings

Here is where the study gets more nuanced. Despite the high theoretical exposure scores, there is little evidence to date that AI is actually responsible for displacing workers. Companies often cite AI when announcing layoffs, but isolating AI as the real driver behind those cuts is difficult.

The clearest warning signs involve young workers. The study's authors found suggestive, anecdotal evidence that entry-level positions are shrinking. Cerullo spoke with a senior programmer who said his firm is hiring fewer junior coders. The reason? He uses AI as if it were a junior programmer, handling tasks that would previously go to a less experienced colleague.

This creates a potential pipeline problem. Junior roles are how workers gain the experience needed to become senior professionals. If those entry points shrink, the long-term effect could be a shortage of experienced workers down the line.

---

Gen Z pivots toward skilled trades

The uncertainty around white-collar job security is already changing how young people think about careers. A survey cited in the segment shows that 77% of Gen Z say it is important that their future job is difficult to automate.

That concern is driving renewed interest in skilled trades like electrician, plumber, and mechanic. These are jobs that require physical dexterity and on-site presence, qualities that keep them firmly in the "least exposed" category. For a generation that grew up hearing about the promise of knowledge work, the pivot is notable.

---

How to interpret these findings

There are several things worth keeping in mind when reading Anthropic's research.

First, Anthropic is studying the impact of its own product. That is unusual. Most AI companies avoid publishing estimates of how many jobs their technology could affect. Whether this signals genuine transparency or strategic framing is an open question. Publishing the research lets Anthropic shape the narrative around responsible AI development.

Second, theoretical AI exposure is not the same as actual job loss. A job where AI can do 75% of the tasks does not mean 75% of those workers will lose their jobs. It could mean each worker becomes more productive, or that teams get smaller gradually over years.

Third, the study's finding that job losses are "not showing up in the data just yet" does not mean they will not. AI adoption is still early. The gap between what AI can do and what companies actually deploy gives workers a window to adapt, but that window may narrow as adoption accelerates.

---

Glossary

---

Sources and resources