A Score of 60 Is Not a 60% Chance Your Job Disappears

The short answer: a score of 60 means roughly 60% of the hours you reported land on tasks the exposure research rates as ones a language model, or software built on one, could do at least twice as fast without losing quality. That is a description of your week. A probability of job loss is a claim about an event happening to you. The tool measures the first and has nothing to say about the second.

Two claims about different objects

A probability is a statement about an event: this thing will happen with this likelihood. The composition of your week is a statement about a quantity: this share of your hours has this property. They are not convertible into each other. Knowing that 60% of your hours are on exposed tasks tells you nothing, by itself, about whether any event follows, any more than knowing 60% of a building's floor space is carpeted tells you whether it will be renovated.

The Science rubric that the ratings come from is explicit that it measures exposure, defined as a time reduction on a task, and that exposure is not displacement. The AI Job Risk Check score inherits that definition and does not add a forecast on top.

What fills the distance

The distance between the composition of your week and any event in your career is filled with things no exposure rating touches:

• Whether the technology gets bought by your employer, and when
• Whether it is legally allowed to do the task without a licensed person
• Whether the output is good enough at your quality bar, rather than an average one
• Whether demand for the whole occupation grows faster than the tasks shrink
• Whether the absorbed tasks were what your employer was paying you for

I have written about each of these in what an AI job risk score cannot know. The point here is simpler: every one of them sits between the measurement and the outcome, and the score does not model any of them, because it cannot.

Why I will keep drawing this line

Tools in this category have a strong incentive to report a probability, because a probability sounds like a prediction someone stood behind. The published research does not support one. Nobody has a validated model that takes task exposure and returns the likelihood that a specific person loses a specific job, and anyone reporting one has either invented the conversion or borrowed a number from somewhere that did.

So the tool reports what was measured and stops. The score is the share of your hours on exposed tasks. The report shows which tasks, with the rating on each. Where the hours might go next is a conversation, and it is yours.

How to read a 60

Read it as: about six in ten of my working hours are on tasks the technology can already reach. Then ask two questions. First, which tasks are those, and are they the ones I want to be doing anyway? Second, of the five items above, which one is doing the most work in my situation? For someone in a licensed profession it is usually regulation. For someone in a growing field it is usually demand. For someone at a slow-moving employer it is usually adoption. Naming the dominant one turns a number into a plan.

If you have seen a tool that quotes a probability

Ask what it cites. If the answer is the Science paper, the tool has converted an exposure share into a likelihood without a method for doing so. If the answer is a proprietary model, ask what outcome data it was validated against, because the outcome data for AI displacement at the individual level does not yet exist in any form I have found. The honest answer to "what is the probability I lose my job to AI" is that nobody knows, and a tool that claims otherwise is telling you something about itself.

Frequently asked questions

What does an AI Job Risk Score of 60 mean?

That roughly 60% of the working hours you reported fall on tasks the exposure research rates as ones a language model or software built on one could do at least twice as fast without losing quality. It describes your week, not a likelihood.

Is the score a probability of losing my job?

No. A probability would require modelling adoption, regulation, quality standards, demand, and employer priorities. The score measures none of those, and no published research supports converting exposure into a displacement probability.

Why does the tool not report a probability?

Because it would be a number made up rather than measured. The score reports the composition of your week, which is what the underlying research supports, and stops there.