No black box. Here's exactly how the AI Job Risk Test measures exposure, the research it stands on, and an honest account of what it can and can't tell you.
The single most important principle is this: AI acts on tasks, not job titles. Two people with the identical title can face very different exposure, because they spend their days doing different things. A "marketing manager" who mostly writes routine copy is in a very different position from one who mostly manages relationships and makes strategic calls.
So the test never scores your title. It scores the work you actually do, and rolls that up into a single exposure number. This is the same task-based approach used in serious economic research on automation, and it's why the test asks about your day-to-day activities rather than just asking what you're called.
Every task you do sits somewhere on five axes. Together they capture what makes work easy or hard for AI to take over. Your answers are scored across all five:
Repetitive, rule-based, predictable work, the same steps most days, is what AI does best. Work that changes constantly and follows no fixed pattern is much harder to automate.
Work that happens in the unpredictable physical world, using hands, tools, and bodies in changing environments, remains very hard for AI and robotics. Work that happens entirely on a screen is far more exposed.
Work whose whole point is the human relationship, care, persuasion, negotiation, comfort, is protected, because people want a person. Work where no relationship is involved is easier to hand to a machine.
Fresh judgment on genuinely new situations, and original creative work, resist automation. Applying the same known rules to familiar cases does not.
When someone must be legally or professionally accountable, a licensed sign-off, a safety-critical call, a decision with serious consequences, a human stays in the loop by necessity, even when AI assists.
Your responses across those five dimensions are weighted and normalized onto a single 1 to 100 scale, where 1 means most resilient (the work is almost all things AI is bad at) and 100 means most automatable (the work is almost all things AI is good at).
A low score doesn't mean "nothing will ever change", it means the core of your work is hard to automate. A high score doesn't mean "you're replaceable tomorrow", it means a large share of your current tasks are the kind AI takes first. Most jobs land in the middle, in the augmentation zone, where AI changes how the work is done without eliminating the role.
The task-based framework draws on the body of economic research that studies automation seriously, including:
Full titles and authors are given so you can look up the original research directly. These sources inform our task-based approach; Moroporo's scoring model and results are our own.
The through-line across all of it is the same principle this test is built on: automation acts on tasks, and a job's exposure depends on its particular blend of them.
The score reflects typical exposure for the pattern of work you describe. It is a well-grounded starting point for thinking, not a forecast of your specific future.
Being straight about what the test can't do is part of being trustworthy:
We'd rather tell you the honest boundaries of the tool than oversell it. A score you can trust is worth more than a score that pretends to be certain.
Each job page on this site shows a typical score for that role. But an average smooths over the very differences that matter, two people with the same title can land far apart. The test exists to give you a number based on your actual task mix, plus a plan built around it.
Score your own role, free →