How Our AI Job Risk Score Is Calculated (Every Weight Shown)

Most AI job risk tools give you a percentage and no way to check it. This post does the opposite for ours. Below is every number the AI Job Automation Risk Calculator uses, the order it applies them in, and worked examples you can reproduce yourself.

One thing to know before the tables: the weights are our own estimates. They are not taken from a dataset, and no study assigns a score to an industry or a job title the way we do. The score is a structured way to think about your work, and the sections on sources and limits below say plainly where it is weak.

The Score Is a Sum

The calculator starts from a number for your industry and then adds or subtracts points four times:

The total is rounded and held between 0 and 100. We call the result task exposure: a rough measure of how much of your daily work AI tools could plausibly take on or assist with. It is not the chance of losing your job.

Step 1: Your Industry Sets the Starting Score

Each of the twelve industries has a fixed starting score. The gap between the highest and the lowest is 45 points.

Step 2: Words in Your Job Title Move It

The calculator does not hold a list of occupations. It looks for certain words inside whatever title you type and adjusts the score for each one it finds.

Words That Raise the Score

Words That Lower the Score

Two details matter here. Adjustments stack, so "Data Entry Clerk" collects the points for both "data entry" and "clerk". And a title with none of these words, such as "Accountant" or "Nurse", gets no adjustment at all.

Step 3: The Tasks You Tick

You choose three to five tasks from a list of nine. Each one adds or subtracts a fixed amount.

Ticking every task that raises the score adds 36 points. Ticking the five that lower it removes 44. That is a range of 80 points, against 45 for the choice of industry, so your answer to this question usually matters more than the industry you pick.

Step 4: Experience and Education

These are the smallest adjustments, and both only lower the score.

• Experience: one point off for every two years, up to a maximum of 10 points at 20 years.
• Education: 8 points off for a doctorate, 5 for a master's degree, 3 for a bachelor's degree, 2 for vocational or trade school, and none for high school or "other".

Worked Example: Three Titles in One Industry

Take three people in Finance & Banking, each with 20 years of experience. The clerk ticks data processing, report generation and filing. The analyst ticks data processing, report generation and strategic planning. The director ticks leadership, strategic planning and report generation.

All three start from the same industry figure. The clerk and the director end up 78 points apart, and most of that gap comes from the title words and the tasks, not from experience or education.

Worked Example: One Title, Three Kinds of Day

Now hold everything still except the tasks. An accountant in Finance & Banking with a bachelor's degree and 20 years of experience gets three different scores depending on what the working day looks like.

Same title, same industry, same background, and a 52-point spread. This is the most useful thing the calculator shows: a job title says little about exposure, and the mix of work inside it says a lot.

What the Bands and the Other Lines Mean

The score is sorted into four bands: low (30 or under), moderate (31 to 60), high (61 to 80) and very high (above 80).

Two other lines on the result card come straight from the score and add no new information:

• The timeline sentence is one fixed sentence per band. Everyone in the high band sees the same years.
• The "higher or lower than X% of jobs" line is a lookup with six steps. It is not a comparison against a survey of real jobs, so read it as a restatement of your band.

Where the Numbers Come From

We set the weights by judgment. The direction they point in follows published research on which kinds of work are exposed to generative AI, but the size of each number is ours.

The clearest agreement is on clerical work. A 2023 working paper from the International Labour Organization, Generative AI and Jobs, found that clerical work was the only broad occupational group with high exposure: 24 per cent of clerical tasks were rated highly exposed and a further 58 per cent medium. In every other group, highly exposed tasks made up between 1 and 4 per cent. The authors concluded that the larger effect is likely to be the automation of some tasks within a job, leaving the job itself in place. That matches Admin / Clerical having our highest starting score.

There is also a point where research pulls against our model. GPTs are GPTs, a 2023 study by Eloundou, Manning, Mishkin and Rock, estimated that about 80 per cent of US workers could have at least 10 per cent of their tasks affected by large language models, and about 19 per cent could have at least half affected. It found the effects at every wage level, with higher-income jobs potentially more exposed. Our calculator lowers the score for seniority and for education. That reflects a view that judgment and accountability are hard to hand over, but it is an assumption, and that study is a reason to hold it loosely.

What the Score Cannot Tell You

• It reads words, not occupations. A title is matched against a short list of words. Two people doing the same work under different titles can get different scores, and an unusual title may match nothing.
• Nine tasks cannot describe a job. Each task is a broad category with one fixed weight. "Data processing" covers work that AI already does well and work it does badly.
• Scores pile up at the ends. Because the parts are simply added, strong combinations run past the limits. A "Data Entry Clerk" in finance with three routine tasks totals more than 100 before the cap is applied.
• It knows nothing about your employer or country. Whether a task is actually automated depends on cost, regulation and what your organisation decides, none of which the calculator sees.
• Exposure is not job loss. A high score means much of your current task mix is the kind of work AI tools are being built for. What happens to the role is a separate question.

How to Get Something Useful From It

1. Run it twice. Once with the tasks that fill your week now, and once with the tasks you could realistically move toward. The difference between the two scores is more informative than either number.
2. Look at the breakdown, not the headline. Using the tables above, work out which step contributed most to your score. If it is the tasks, that is the part you have some control over.
3. Treat the number as a prompt for a conversation. It is a reasonable starting point for talking with a manager about which parts of your role are likely to change.

For the reasoning behind looking at tasks instead of whole jobs, see Is Your Job Safe from AI? and Which Parts of Your Job Will AI Change First?