Career & Jobs
How to Read an Automation Exposure Score Without Panicking
By Jim Vernon, Editor, AI Intelligence International · Published 8 January 2026 · Reviewed against our editorial standards · About the author
Automation exposure scores have become the first thing people search for when a new model ships. They are useful, but only if you understand what the number is doing. An exposure score is not a probability that you will be fired. It is an estimate of how much of the work you produce in a normal week could be drafted by software that already exists.
That distinction changes every decision you make next. A high score on a task-heavy job means your day is about to be rearranged. It does not mean the role disappears, and it certainly does not mean it disappears this year.
Key takeaways
- What the number is actually measuring: Every job decomposes into task families: producing routine text, moving structured data between systems, judging incomplete information, coordinating people, and physically handling objects or patients or machines.
- Why two people with the same title score differently: Job titles are administrative artefacts.
- Compression happens long before replacement: The observable pattern in exposed fields is not mass dismissal.
- The four decisions worth making after you see your score: First, move up the accountability ladder.
What the number is actually measuring
Every job decomposes into task families: producing routine text, moving structured data between systems, judging incomplete information, coordinating people, and physically handling objects or patients or machines. A model scores each family separately and then weights them by how much of your week they consume.
Routine drafting scores highest because a language model can produce an acceptable first version in seconds. Structured data handling scores nearly as high, because the same models are now reliable at classification when the categories are well defined. Judgement under accountability scores far lower, not because software cannot attempt it, but because organisations rarely let it sign off unsupervised.
So a 78 does not mean you are 78 percent replaceable. It means roughly three quarters of your measurable output falls in task families where a first draft is now cheap.
Why two people with the same title score differently
Job titles are administrative artefacts. Two marketing coordinators can have almost no tasks in common: one writes forty emails a week, the other runs events and manages agencies. The first is heavily exposed, the second barely at all, and both would type the same title into a calculator.
This is the single biggest source of confusion in AI-risk coverage. When a headline says a profession is 60 percent exposed, it is describing the average of a distribution that is often bimodal. Read your own week, not the profession's average.
A practical exercise: list the ten things you did last week that someone paid for. Mark each as drafting, data movement, judgement, coordination, or physical. The proportion that lands in the first two buckets is your personal exposure, and it is usually more accurate than any title-level score.
Compression happens long before replacement
The observable pattern in exposed fields is not mass dismissal. It is compression: the same volume of work handled by a smaller team, each member reviewing machine output rather than producing every artefact by hand. Compression shows up as a hiring freeze, a wider span of control, or a backfill that never happens.
That matters because the people most affected are the ones who would have been hired next year, not the ones already in the seat. If you are employed and mid-career, your realistic risk is stagnation and a thinner internal job market, not a sudden exit.
It also means the defensive move is not to work faster. Working faster inside an exposed task family simply makes the compression maths more attractive to whoever is running the budget.
The four decisions worth making after you see your score
First, move up the accountability ladder. Own the review step, the client relationship, or the definition of what good output looks like. Those remain human responsibilities long after the drafting is automated, because someone has to be answerable when the output is wrong.
Second, get deliberately good at directing the tools that threaten your task list. The people who survived spreadsheet automation were the accountants who learned spreadsheets, not the ones who avoided them.
Third, build a portfolio of work that shows judgement, not volume. Three case studies explaining why you chose an approach are worth more in an interview than three hundred pieces of routine output.
Fourth, put a date on it. Reskilling with no deadline is a hobby. Pick a quarter, pick one capability, and measure whether your week actually changed.
When a high score should worry you
There are conditions that turn a directional score into a real warning. If your output is fully digital, easy to evaluate against a clear standard, produced in high volume, and not customer-facing, every ingredient for fast automation is present.
Add one more signal: if your employer has already deployed a model into an adjacent function and reported savings publicly, the internal appetite exists. Deployment is contagious inside a company because the political cost of the first project has already been paid.
Conversely, if your work involves regulated sign-off, physical presence, or negotiation with people who care who they are speaking to, the score overstates your risk considerably.
How to use the calculators on this site together
Start with the exposure score to get the band. Then run the replacement timeline to see how the same task mix translates into years rather than points, which is the more useful planning horizon.
Follow with the safe-skills check to find which parts of your current role are actually defensible, and the reskilling path if the answer is uncomfortable. Together they turn a number into a quarter's worth of concrete action, which is the only thing a score is good for.
Frequently asked questions
Is an exposure score a prediction that I will lose my job?
No. It estimates how much of your weekly output a current model could draft. Historically that translates into slower hiring and wider spans of control before it translates into dismissals, and many exposed roles change shape rather than disappear.
Why did my score change when I reworded my job title?
Because the score reads the task mix implied by the words. 'Content writer' and 'content strategist' imply different proportions of drafting versus judgement, so they land in different bands. Use the wording that describes your actual week.
Should I tell my manager about my score?
Only in the form of a proposal. A score alone invites an unhelpful conversation. A score plus a suggestion for which parts of the workflow you want to own is a career conversation worth having.
How often should I re-run it?
Twice a year is enough. Model capability moves faster than that, but organisational adoption does not, and adoption is what determines whether your week changes.