Ask most people which jobs are safest from automation and they will picture someone in a hard hat or a hospital, and someone in an office as the one at risk. Our data says the opposite is closer to the truth.
We scored 427 professions across 127 countries for automation exposure, using Frey & Osborne task probabilities weighted by each country's AI readiness and employment trend. The average profession sits at about 14% exposure. The interesting part is the two ends of the list.
The safest work is physical, not prestigious
The lowest-exposure jobs are almost all manual and physical: hunters and trappers, bricklayers, building cleaners, and a long tail of labouring roles all land near 4% exposure. These are jobs that happen in unpredictable physical space, and that is still the thing machines are worst at. Prestige and pay have very little to do with it.
The most exposed work is desk work
At the other end sit the roles we tend to think of as "safe office jobs". Data Entry Clerks top the list at 32.6% exposure, followed by contact-centre sales, typists, and a wall of finance and accounting clerks. Routine that lives inside a screen is exactly what current AI automates first.
One name on the exposed list should make tech workers pause: web and multimedia developers show up at 28.2% exposure. Writing code is not the safe harbour it was assumed to be.
Where you are changes the answer
Exposure is not fixed by job title alone. The same profession carries a different risk depending on the country's AI adoption and labour market, which is why the map below ranges widely across the 127 countries we cover. Two people with identical titles in different places do not face the same future.
Automation exposure: 10 lowest vs 10 highest of 427 professions
Average automation exposure by country
So what
A job's safety today is a poor guide to its safety in five years, and "high-skill" is not the same as "safe". The useful move is not to chase the lowest-risk job on the list, it is to understand which of your specific tasks are exposed and which are not, then shift your weight toward the ones machines handle badly. That is what your Career Health read is for.