Five Signals Your Job Is AI Exposed, Read From One Ordinary Week

Is my job at risk from AI? The honest answer is that your job title cannot tell you, because two people holding it can be in completely different positions. What can tell you is your own calendar. Below are five signals, each readable from a single ordinary week, and each pointing at something specific you can change.

Run this against a real week rather than an idealised one. Pick last week. Not the week you had a big presentation, and not the quiet one in August.

Your work arrives already specified

Somebody hands you a brief. The brief is clear. You produce the thing the brief asked for, and the main question anyone asks afterwards is whether you produced it well.

This is the single strongest signal, because a stated task with a checkable output is precisely the shape current systems handle best. The exposure is not about your competence. Excellent execution of a fully specified task is exactly what gets compressed first.

The inverse is the safer position: work where the first hour goes into deciding what the brief should have said.

There is one correct form of the output

Ask yourself whether a second competent person, given the same inputs, would produce something recognisably the same. If the answer is yes, the work has a canonical form, and canonical forms are learnable.

Standard reports, routine correspondence, formatted summaries, first-pass code from a clear spec, translations of unambiguous text. None of these are trivial to do well. All of them have a target that can be described, which is what matters here.

Where two good people would reasonably diverge, and where the divergence is the value, exposure drops sharply.

Nobody asks you what the goal should be

Track how often, in one week, somebody asks your opinion on what to do rather than how to do it. Include the small instances. A client changing direction, a colleague checking whether an approach is worth pursuing, a decision you escalated or absorbed.

If that count is zero, you are positioned downstream of every decision in your process. Downstream is where specified work lives, and specified work is where the pressure lands. This is the same distinction that makes job title calculators unreliable, which we cover in how AI job risk scores are calculated.

Your week contains no moment of noticing

This one is subtle and it is the one people score wrong.

Somewhere in a good week, something should have failed to add up. A number that looked off. A brief that no longer matched what the client had said two calls ago. A quiet colleague whose silence meant something. A plan that was still technically on track while quietly becoming the wrong plan.

Noticing is the part of work that never appears in a task list, which means it never appears in an automation estimate either. If you genuinely cannot find one instance in your week, that is worth sitting with. Either the role has no surface for it, or the noticing is happening and you are not crediting yourself for it.

The judgement calls belong to someone else

When something ambiguous happens, does the call route to you, or through you to somebody senior?

There is no shame in the second. Early in a role it is correct. But it locates you precisely: you are handling the specified part while somebody else holds the unspecified part, and the unspecified part is the durable one. The six domains where that judgement is measured are set out in the skills AI can't replace.

So, is my job at risk from AI?

Count your signals.

Zero to one. Your work is mostly unspecified, which is the resilient position. The risk here is different and worth naming: it is that you never build the technical fluency to work alongside these tools, and get out-executed by someone who has both.

Two to three. The common case. Part of your week is specified and compressible, part of it is judgement. The useful move is to notice which meetings and which tasks fall on which side, and deliberately move your hours toward the second.

Four or five. Worth acting on this quarter. The instinct is usually to learn a tool, and that is the smaller half of the answer. The larger half is to get closer to where the goals are set. Ask for the ambiguous work. Volunteer for the piece nobody has scoped. Be in the room where the brief is written rather than the room where it is executed.

None of this is a prediction about you. It is a description of where you currently sit, which is the only thing a checklist can honestly give you.

If you want the same read produced from your answers rather than your own estimate, the check scores six domains of judgement separately and takes about seven minutes.