Will AI take your job? Read the data, task by task.
Below, every dot is one of 785 real occupations, placed by its measured overlap with what AI actually does today. Find yours, see which of your tasks AI can and cannot do, and leave with a plan. No doom meter. No fake precision.
785 occupations covered. Job title not on the list? Assess it from your actual tasks
Jobs don't get automated. Tasks do.
The studies disagree about the numbers, and the disagreement is mostly about the unit. Count whole occupations and you get 47%; count the tasks inside them and you get 9%. This site counts tasks, and the guide on tasks versus job titles shows why that changes the answer five-fold.
-
Your job becomes its activities
Each occupation is decomposed into the work activities that actually fill its week, weighted by importance, using the U.S. Department of Labor's O*NET framework.
-
Each activity gets a measured value
Microsoft Research measured how often AI is really used for every activity, how successfully it completes it, and how much of it AI covers. Observations, not predictions.
-
The overlap becomes your move
You see which of your tasks overlap with AI and which don't, adjust the estimate for how you personally work, and leave with a concrete plan: what to shift toward, what to own, what to learn.
Where 785 occupations actually land
Bands are quintiles of the measured score distribution, so each band holds one fifth of all occupations. A band tells you where you stand relative to everyone else, not a percentage chance of losing your job. Nobody can honestly give you that number.
How exposed is your field?
Every occupation belongs to one of 22 fields. Each field page ranks all of its occupations and shows the band mix, and the links work with or without JavaScript.
The edges of the distribution
Highest AI applicability
Lowest AI applicability
Browse and filter all 785 occupations ยท Compare any two side by side
High exposure is not a pink slip
Exposure means overlap, not replacement. A high score says AI can already do meaningful parts of your job's activities. History cuts both ways, and the whole of it matters: ATMs did not eliminate bank tellers for two decades, they changed what tellers did, and then teller employment fell by a third between 2012 and 2022. The prediction track record tells both halves. Anthropic's usage data put augmentation slightly ahead of automation on consumer Claude in its January 2026 report, 52% against 45%, though that split has reversed between reports and runs the other way in API traffic. Exposure, automation and replacement sets out what the measure can carry.
The labor market is churning, not vanishing. The World Economic Forum's 2025 employer survey projects roughly 92 million jobs displaced and 170 million created by 2030, a forecast of employer expectations whose three expired predecessors had no published retrospective we could find in August 2026. The IMF estimates about 40% of jobs globally are exposed to AI, and stresses that roughly half of those may be complemented rather than replaced.
But "someone using AI" is a real threat. The consistent pattern across studies: the near-term risk isn't a robot taking your seat. It is a person in your field who uses AI well, producing two or three times your output. Exposure data tells you exactly which of your tasks that person will accelerate.
That is why every page here ends with a plan. Which tasks to hand to AI first, which human-anchored tasks to double down on, and which adjacent roles share your skills with less exposure. All of it computed from real activity overlap, not vibes.
Job title not on the list? Freelance? Hybrid role?
Build your exposure profile from what you actually do all day: ten task families, measured against the same activity data.
Questions people actually ask
Is this "AI-powered"? What actually computes my result?
No model runs when you use this site, and nothing you type leaves your browser. The scores were computed by Microsoft Research from 200,000 anonymized Bing Copilot conversations and published openly (CC BY 4.0). We turn that dataset into readable, task-level pages and transparent arithmetic. Every step is documented on the methodology page.
Does a high score mean I'll lose my job?
No. It means AI can already perform meaningful parts of your occupation's activities, which historically leads to job redesign, wage pressure on routine work, and advantage for people who adopt the tools early. Displacement is one possible outcome among several, and the studies that predicted mass unemployment on a timetable have so far been wrong.
Why should I trust this over other "will a robot take my job" sites?
Three reasons. First, the data is measured usage, not a 2013-era expert guess about "computerisation probability." Second, we show our work: every occupation page lists the underlying activities and their measured AI overlap. Third, we never print fake precision. You get bands, percentiles and ranges with stated limitations, not "87.3% doomed."
Is my data collected?
No accounts, no cookies, no tracking scripts, no analytics beacons. Quiz answers are processed in your browser and go nowhere. The one optional convenience, the "recently viewed" list, lives in your browser's local storage and never leaves your device.
My country isn't the U.S. Does this apply to me?
The occupation taxonomy and activity weights are U.S.-based (O*NET), but the task content of an accountant's or nurse's day is broadly similar across developed economies, and the AI usage data itself is global. Treat the score as a task-exposure signal, not a national labor-market forecast. Local regulation, wages and language markets shift the timeline.
How current is the data?
Scores come from Microsoft Research's 2025 "Working with AI" study, the largest published task-level measurement as of August 2026, joined with the O*NET 30.0 database (2025). We revisit the sources and refresh when materially better data is published. The current data vintage is always listed in the methodology.