JobRiskAI

AI and your job, explained plainly

The occupation pages give you a number. These pages explain what that number is, what it is not, and what the evidence actually supports. Every figure is sourced and dated, and the honest answer is usually in the middle: neither "AI takes everything" nor "AI changes nothing".

If you have just looked up an occupation, start with how to read your score, then exposure, automation and replacement. If you already know your score is high and want the practical part, go straight to what to do if your score is high. If you are sceptical of the whole exercise, which is reasonable, start instead with how accurate past predictions have been. The other three fill in the evidence behind these four.

How to read your AI exposure scoreWhat the number on your occupation page actually measures, and the three ways people misread it. Exposure, automation and replacement are three different thingsWhy a high score is not a prediction, and what the labour market data actually shows so far. Why AI risk is measured in tasks, not job titlesCounting whole occupations gives 47%. Counting tasks gives 9%. Why the unit changes the answer five-fold. What the usage data shows AI is good and bad at, at workThe pattern across 332 work activities, and why creative work splits rather than moving as one block. Hands-on and licensed jobs: what the data shows and what it does notManual trades score low. Licensed professional work does not. Why the two are often wrongly lumped together. How accurate have AI job-loss predictions been?Where the forecasts can be graded, the headline readings failed, but the two retrospectives point in opposite directions. What to do if your exposure score is highWhat the evaluation evidence says actually helps, which is not what most advice assumes.
How these pages are written. Each one states its answer first, then shows the evidence behind it. Figures come from named primary sources (research papers, statistical agencies, published datasets) and are quoted with the caveats those sources state themselves. Where the evidence is genuinely contested, the page says so instead of picking the more dramatic side. The scoring used across this site is documented separately on the methodology page.