"Safe From AI" Is Not the Same as "Safe"

Our site includes a ranking of jobs that are harder for AI to automate. As the people who publish it, we think how it is read matters a great deal. Finding a lower-scoring occupation and reading it as "I am safe" is the interpretation that probably deserves the most caution.

Plumbers and nurses, for instance, tend to sit on the less-exposed side in our assessment. Work that only begins on site, where physical action and judgement are bound together, is exactly what today's AI is weakest at — and that is unlikely to change quickly.

"The occupation survives" is different from "the work stays the same"

A nurse's documentation, a plumber's quoting and ordering, a teacher's grading and lesson preparation — even occupations on the less-exposed side contain routine tasks that AI is good at. When those parts are absorbed, the occupation itself remains, but what the occupation asks of a person may shift. A nurse whose documentation is automated may be asked to handle more patient contact; a plumber whose quotes are produced faster may find that scheduling and proposals become where differences show.

So a less-exposed assessment is closer to "even when change comes, the core value of this occupation is relatively more likely to remain" than to "no change is coming." And a higher score, likewise, may be less "the occupation disappears" and more a clue that a part once thought core turns out to be routine.

The more useful question may not be "which job is safe," but "which parts of my work are likely to still be done by people in five years."

There are second-order changes a score cannot capture

One thing that is easy to overlook is the pressure of people moving into less-exposed occupations. If clerical roles shrink and workers move toward fields like care or construction, that movement itself can affect conditions and ways of working. Even without being directly automated, a role can change indirectly through the wider labour market — and a score cannot capture that second-order effect.

For that reason, a "safest jobs" ranking may be better read not as a list of places to shelter, but as material for thinking about which kinds of skills are likely to have a longer shelf life.

In the end, the more useful question may not be "which job is safe," but "which parts of my current work are likely to still be done by people in five years, and what can I do to build those parts up." We hope the ranking is useful as a starting point for that.