How do I tell if my job is at risk from AI?

Why the published lists are close to useless for you

Occupation-level risk lists average across everyone holding a title. You are not an average. Two people with the same title, at the same company, can carry completely different exposure based on which parts of the work they have drifted toward over the years.

The lists also tend to describe what is technically possible rather than what an employer will actually do, and those diverge for years at a time. What closes the gap is not capability, it is a budget cycle.

Signal one: what your employer has stopped doing

Organizations telegraph this long before they announce it, and the tells are absences rather than events, which is why they are easy to miss.

1. 01A colleague in a role like yours left and was not replaced, and the work was absorbed rather than rehired.
2. 02A req for your kind of role sat open for a quarter and then quietly closed.
3. 03Your team's work is increasingly described in terms of output volume rather than judgment or ownership.
4. 04Tooling decisions that affect your workflow are being made without your input, by people who do not do your job.
5. 05New work arrives already scoped, where it used to arrive as a problem to define.

Signal two: how much of your week arrives pre-specified

Take your last two weeks and sort each recurring task by whether it was fully specified before you began. Not 'was it hard' — whether the definition of done existed in advance.

Work that arrives specified is work someone has already reduced to a brief. Once a brief exists, the question of who or what executes it is an economic one rather than a technical one, and the answer tends to move over time.

Work that arrives as a problem to define is the opposite. The value is in producing the brief, and that has stayed stubbornly difficult to hand off.

If most of your week arrives pre-specified, your exposure is high regardless of your title, your tenure, or how well you are regarded. That last part is the one people resist, and it is the one that matters.

There is no safe. There is only leverage.

The question people want answered is whether they are safe. It is the wrong target, because safety implies a stable state, and no current role is stable in that sense.

The useful target is leverage: whether what you supply is scarce in the room you are standing in, and whether you could demonstrate it to a different room on short notice. Leverage is portable. Safety is not.

This reframing has a practical consequence. It means the work of becoming harder to replace is the same work as becoming easier to hire, so it pays off whether or not the risk ever materializes.

The advantage of moving while you still have a job

Everything above is easier to act on while employed, and this is the part worth being blunt about.

You can have exploratory conversations without urgency in your voice. You can be selective. You can build evidence of the judgment you supply while you still have a place to supply it. You can take the slower route if it is the better one.

None of that is available thirty days after a layoff with a runway counting down, which is when most people start. The window is open precisely when it feels least necessary to use it.

Where this stops being enough

These signals describe exposure, not a timeline. A high-exposure role can persist for years on budget inertia, and a low-exposure one can disappear in a reorganization that has nothing to do with technology. Treat this as a reason to build leverage rather than a prediction about a date.