FAQ
Frequently asked questions
What the % measures, what it does not, and where each number comes from.
The long version is on Method.
What does the percentage measure?
Human time that can already be saved with AI or a robot, at an acceptable result.
If a task takes 100 minutes and today’s tools leave 20, that task is 80%. The occupation is the sum of those tasks, each weighted by its share of the week.
What does it not measure?
It is not the odds you will be fired. It does not estimate how many jobs disappear. It does not speak to your contract.
The percentage is about the occupation, not your seat, your firm or your bargain. A two-minute spellcheck does not make a job 80%.
How is a job split into tasks?
Tasks come from O*NET. One page = one O*NET code. “Junior developer” is Computer Programmers (15-1251.00), not software developers in general.
O*NET does not publish what share of the week each task takes. Importance (IM) is not time; we do not use it as weight. Frequency (FT) says how often, not how long. We turn that into relative time and normalise to 100%. That is our estimate, not an “O*NET weight”.
How is the score calculated?
score = Σ (weight × saving). Weights add up to 100% of the week. Saving on each task is estimated separately and is not retouched to fit the headline.
Nobody types the percentage. The number on the page is that sum, rounded. The long version is on Method.
What does Public evidence mean?
There is a source you can open: O*NET, ESCO, a benchmark (SWE-bench, LiveCodeBench…), a dataset, a paper or a statistic.
If a task is public, it must carry a URL. On the page: “Public” on the task, and “See evidence” when there is a link.
What does Estimate mean?
No source is precise enough to pin that task’s saving. The method still estimates it: what share of the time current AI can already save, at the same quality.
Public + estimate = 100% of the week. Almost every page today is estimate (low confidence). It is not a scientific probability.
Where do the sources come from?
O*NET and ESCO describe the work (tasks, skills, codes). Benchmarks and papers say what AI can already do.
Each page shows the O*NET and ESCO/ISCO code when we have it. First measured 25 Aug 2026.
Is the percentage about my job, or the occupation?
The occupation. Two people with the same title can have different weeks. We measure the occupation, not your calendar or your manager.
Why aren’t all jobs in the ranking yet?
The ranking and directory publish 10 pages a day, from 26 Aug 2026. The rest are measured and queued, but they do not show in those lists until their day.
The analyser on the home page does know the queued pages: if you ask for “agricultural engineer”, you get that page. Search engines do not index them until they go live.
I type “engineer” or “programmer” and several pages appear. Is that a bug?
No. Those names cover many occupations. Instead of picking one at random, we list the matches so you can choose the one that fits: junior developer, senior, PLC programmer…
If you type the full title (“agricultural engineer”), you go straight to that page.
What is confidence (high, medium, low)?
High: most of the week’s weight has recent public evidence. Medium: a mix. Low: almost all estimate.
Most pages today are low. This is not a paper’s confidence interval.
AI and robot: the same thing?
No. The headline is the combined figure: time already saved with AI or with a robot. Where we split it, digital is software/AI and robot is a physical machine. A welder is not an accountant.
How often is this updated?
First measured 25 Aug 2026. When a task’s saving changes, it will carry a date. The page score is recalculated with the formula; the headline is not typed by hand.
Can I ask for a job that isn’t here?
Yes. If the analyser finds no page, press “Please measure it”. We keep that short phrase (80 characters max, no CVs) to know what to measure next.