Jobs safe from AI? The occupations least exposed to generative AI

"Which jobs are safe from AI?" is one of the most searched questions about work and technology. The honest answer is that no list can promise safety: tools change quickly, and even jobs AI cannot do are changed by the AI around them. What the data can show is which occupations are least exposed to generative AI today, and which roles employers expect to grow. This page sets that out, using the ILO's 2025 exposure index and the WEF's 2025 employer survey.

The short answer: the occupations least exposed to generative AI are mostly hands-on, in-person and care jobs: construction trades and labourers, cleaners, farm workers, cooks, hairdressers, electricians and plumbers, along with teachers, nurses, doctors and social workers, whose work depends on physical presence, judgment and human contact. Many of the same frontline and care roles are among those employers expect to grow most by 2030. Low exposure is not a guarantee: it describes the potential of today's generative AI, not robots, wider automation or economic change.

How exposure is measured

The International Labour Organization scored thousands of work tasks from 0 (cannot be automated with generative AI) to 1 (could be done without a human operator), then classified occupations by the average and spread of their task scores. Below the four exposed gradients it places two lower categories: minimal exposure, where some interaction with generative AI may occur without significantly altering the role, and not exposed, where generative AI currently has no observable potential to automate tasks (ILO Working Paper 140, 2025).

The ILO's broader finding explains the pattern: exposure is highest in occupations with the most cognitive tasks and declines as tasks become more physical.

Some of the least exposed occupations

All of these are classed as "not exposed" in the ILO's 2025 index. The score is the average automation potential of the occupation's tasks; for comparison, the most exposed occupation, data entry clerks, averages 0.70 (ILO Working Paper 140, Annex Table A1).

The lowest scores in the whole index belong to physical jobs such as building construction labourers, garbage and recycling collectors, crop farm labourers and vehicle cleaners, all around 0.09.

Linked occupations have their own guide in our will AI take my job? series, which covers 328 jobs with their tasks, sources and outlook. To compare jobs side by side, see jobs most exposed to AI, fastest-growing jobs and their AI exposure, AI exposure by job field, AI exposure by education level, AI exposure by pay and jobs with the most openings.

What these jobs have in common

• Physical work in changing environments: every building site, kitchen and plumbing job is different.
• Human contact at the core: caring for patients, teaching children, counselling people.
• Judgment with consequences: diagnosing, supervising a site, flying an aircraft.
• Accountability that stays with a person: someone must sign off and answer for the result.

Which roles employers expect to grow

In the World Economic Forum's 2025 survey, employers expect frontline roles to see the largest growth in absolute numbers by 2030, including farmworkers, delivery drivers, construction workers, salespersons and food processing workers. Care economy jobs such as nursing professionals, social work and counselling professionals and personal care aides are also expected to grow significantly, along with education roles such as secondary and tertiary teachers (WEF Future of Jobs Report 2025).

Technology roles, such as big data specialists and AI and machine learning specialists, are the fastest growing in percentage terms: being exposed to AI and being in demand are not opposites.

Why "safe" is the wrong word

• Exposure covers generative AI, not all automation. Robotics and other technologies affect physical jobs in ways the ILO index does not measure.
• Scores move. Between 2023 and 2025 the ILO raised exposure for several digital occupations as AI gained new abilities. Future updates may move other jobs too.
• Low-exposure jobs still change. A nurse's charting, a plumber's quotes and a teacher's lesson plans can all be AI-assisted even if the core work is not.
• High exposure is not the same as job loss. The ILO concludes that transformation, not replacement, is the most likely impact for most jobs, because most still contain tasks that need people.
• Worry is widespread anyway. In OECD surveys of finance and manufacturing workers, 60% were worried about losing their jobs to AI in the next ten years (OECD Employment Outlook 2023).

A better question to ask

Instead of "is my job safe?", ask which of your tasks an AI system could do, which need your hands, presence or judgment, and how you would supervise the first kind. That is true in a high-exposure office job and in a low-exposure trade. For the other end of the scale, see which jobs are turning digital collar, and for making the move, digital collar careers.