Which Jobs Are Most and Least Exposed to AI?

In current studies, jobs built around text, numbers and code rank as the most exposed to AI. These are clerical and administrative work, customer service, programming, and parts of accounting and finance. Jobs that depend on physical work or in-person care, such as cleaning, construction, cooking and home health care, rank lowest. Exposure is not the same as job loss. It describes how much of a job's task list AI could perform or speed up, not whether employers will cut the role.

What does AI exposure mean?

Researchers break occupations into tasks and estimate how many of them AI systems could do or assist with. Some measures are theoretical, based on what the technology can do. Others are observed, based on how people use AI tools.

On the theoretical side, the International Monetary Fund estimated in January 2024 that almost 40% of jobs worldwide are exposed to AI. It put the share at about 60% in advanced economies. It judged that roughly half of the exposed jobs in advanced economies could benefit through higher productivity. In the other half, it judged, AI could perform core tasks now done by people. It said that could reduce demand for their labor.

The International Labour Organization's May 2025 index, which covers generative AI (systems that produce text, images or code), arrives at lower figures. It found one in four workers worldwide in an occupation with some exposure and 3.3% of employment in the highest-exposure category. Exposure is uneven. That top category covers 4.7% of women's employment against 2.4% of men's. Total exposure is 34% of employment in high-income countries against 11% in low-income ones. Because most occupations include tasks that still need a person, the ILO judges that changed jobs, rather than lost ones, are the most likely outcome.

Which jobs are most exposed to AI?

• Clerical and administrative work. The ILO finds clerical occupations have the highest exposure. The US Bureau of Labor Statistics (BLS) projects office and administrative support employment to fall 4.0% between 2025 and 2035. That would be a loss of 752,100 jobs. BLS links the decline to automation tools, some of them AI-powered.
• Customer service. BLS expects customer service representative jobs to decline 5% over the same period as more of their tasks are automated.
• Programming and data entry. In research published in March 2026, AI developer Anthropic ranked computer programmers as the most exposed occupation, with a task coverage score of 75%. It put data entry keyers at 67%. The scores are the company's own estimates and draw on usage of its products.
• Information-heavy professional roles. A Stanford study of payroll data places accountants, market research analysts, paralegals and software developers in its most-exposed fifth of occupations. Microsoft researchers analyzed 200,000 conversations with the company's Bing Copilot chatbot in a 2025 paper. They found the highest AI applicability for computer and mathematical, office and administrative support, and sales occupations.

Which jobs are safest from AI?

No job can be called safe with certainty, but work that requires hands, tools or physical presence ranks lowest on these measures. The least-exposed fifth of occupations in the Stanford study includes freight and stock movers, cleaners, janitors, home health aides, construction laborers, dishwashers and plumbers. Anthropic reported that about 30% of workers, including cooks, bartenders and lifeguards, are in occupations with zero coverage. It said that is because their tasks appeared too rarely in its usage data to register.

Low exposure is not a guarantee. Most of these measures concern generative AI software rather than robots, and employment shifts for many other reasons. Separately, BLS projects healthcare support occupations to grow 13.3% through 2035, the fastest of any major occupational group.

Will AI replace software engineers, accountants or teachers?

Official projections do not predict that, though the cases differ.

Software engineers. BLS projects employment of software developers, quality assurance analysts and testers to grow 10% from 2025 to 2035, adding 185,400 jobs. It cites software development for AI as one driver. It projects the narrower occupation of computer programmers to shrink 7% as programming work is automated.

Accountants. BLS projects 5% growth for accountants and auditors. It says some routine tasks, such as data entry, may be automated. It does not expect that to reduce overall demand. It sees advisory and analytical work becoming a larger part of the job.

Teachers. Elementary, middle and secondary school teachers fall in the middle fifth of the Stanford study's exposure ranking. BLS projects little or no change in high school teacher employment through 2035. It points to declining enrollment and state and local budgets as the constraints. Its outlook does not cite AI.

What to watch

Exposure scores are estimates that shift as AI systems and usage change, and they describe tasks, not hiring decisions. Alongside its August 2026 projections, BLS published its own AI exposure categories. They sort occupations into four levels, from low to very high. They use five outside studies, including the Anthropic and Microsoft research. The agency cautions that exposure does not imply job loss. A more direct test is whether employment in high-exposure occupations diverges from the rest.