Your AI Risk Score Tracks Your Week, Not Your Job Title

The short answer: the score is a weighted average, not a count. It asks what share of your working hours land on tasks the research rates as highly exposed. When this reader cut their writing hours, those hours moved to editing, which rates higher than drafting. The hours left one exposed bucket and landed in a more exposed one, so the average rose.

Hours are the denominator

Most people picture the score as a tally: count the exposed tasks, divide by the total. Under that picture, doing less of an exposed task should always lower the number.

The AI Job Risk Check score does not work that way. It takes the exposure rating for each task in your occupation, weights it by the hours you report on that task, and averages. Your working week is the denominator. Every hour has to be somewhere, and the score reflects where.

The task list is fixed. The week is not.

O*NET, the Labor Department's catalogue, holds 18,796 tasks across the 1,016 occupations we score, about eighteen apiece. Nothing you do changes which tasks belong to your occupation. What moves is how your week distributes across them.

This reader is a communications manager. When they cut their reported writing hours, those hours moved somewhere. In their case, to editing and repurposing existing material. Editing rates higher on exposure than original drafting does, because the research treats transforming an existing text as more tractable for a model than producing one from a brief and a blank page.

Two things follow

A score can move without your job changing at all. It tracks your week rather than your role. The same communications manager, in a quarter dominated by stakeholder meetings and event planning, would report a lower score with the same title, the same employer, and the same task list. That is a feature. A number that could not move with your week would be a number about your title, and titles are the wrong unit.

"Do less of the exposed thing" is incomplete advice. The hours have to go somewhere, and where they go is the actual question. Moving ten hours from drafting to editing raises the score. Moving the same ten hours to stakeholder management, live presentation, or negotiating with vendors lowers it. The instruction to cut an exposed task only helps if it comes with a destination.

What this means for reading your own score

When your score comes back, do not read it as a verdict. Read it as a map of where your hours currently sit. Then ask a more useful question than "how do I lower this": if I reallocated ten hours of my week tomorrow, where would they go?

For most people the honest answer is that the ten hours would go to whatever is nearest, which is usually another text task. The reader's editing hours are a case in point. The score is there to make that visible before it happens, so the reallocation can be a decision instead of a drift.

The conversation I want people to have

A score should start a conversation about the shape of the week, not end one about the fate of the job. The communications manager who wrote in now has a clearer view of their own hours than they had before, and a specific question to take to their next planning cycle: which of the non-text parts of this role can I take on more of? That question is worth more than the number that prompted it.

Frequently asked questions

Why did my AI risk score go up when I reduced an exposed task?

Because the score is a weighted average over your hours. If the hours you removed from one exposed task moved to a task with a higher exposure rating, the average rises even though you are doing less of the first task.

Does editing rate higher than writing for AI exposure?

In the task ratings, transforming an existing text generally rates higher than producing one from a brief and a blank page, because current models handle transformation more reliably than open-ended origination.

Can my score change without my job changing?

Yes. The score tracks how your week is distributed across your occupation's tasks. A different mix of hours produces a different score under the same title and employer.