Find your job
Search 923 occupations and 49,839 other job titles, from barista to biostatistician.
See how much of your job AI can already do, how much people actually use it for today, and which tasks robots can handle.
Built on the Anthropic Economic Index, 923 jobs
of the work AI can do at least twice as fast
Very high exposureShare of working time
Superintelligence will not take your job all at once. It takes it one task at a time.
Headlines talk about whole jobs. AI works task by task, so that is how we break it down.
Search 923 occupations and 49,839 other job titles, from barista to biostatistician.
Each task lands in one of four groups: AI is already doing it, AI could do it next, robots can do it, or it still needs a person.
See how much of your week each task takes, then put your learning time where AI cannot reach.
AI can already do at least half the work in 47% of jobs. Real use is far behind: 48% of jobs show none at all yet.
Share of the work AI can do at least twice as fast
AI can do it, people barely use it yet. Real use of what AI can do
Share of working time robots can technically do
One short email a week on how AI is changing work: which jobs are shifting, research worth reading, and what people are doing about it.
The share of a job's working time spent on tasks AI can do at least twice as fast, at the same quality. That includes tasks people already use AI for, and tasks AI can speed up with the right tools, like coding assistants or AI built into company software.
Each task is weighted by how much of the working week it takes, so a task that fills half your week counts far more than one you do twice a year.
How much people actually use AI for this job's tasks today, based on Anthropic's analysis of Claude conversations at work. Tasks where AI does the work count in full and tasks where it assists count as half.
It is almost always far lower than the main score. AI can already do much more than people use it for, and that gap is what is coming next.
Real AI usage comes from the Anthropic Economic Index, which maps millions of privacy preserving Claude conversations to official job tasks. Robot ratings and the time each task takes come from Anthropic's What work can robots do? study from September 2026. For tasks without real use yet, whether AI can speed them up comes from human ratings in the study GPTs are GPTs, the same benchmark Anthropic uses. Job tasks come from O*NET, the U.S. Department of Labor's database.
Yes. It covers one AI assistant, so it misses AI use through other tools like ChatGPT, Gemini or Copilot. Treat it as a floor.
Not on its own. Anthropic found no systematic rise in unemployment among highly exposed workers since late 2022, but it did find suggestive evidence that hiring of younger workers into exposed jobs has slowed.
The score tells you which parts of your work are changing first, so you can decide where to put your energy.
For physical tasks we use Anthropic's 2026 robot study, which rated what today's robots can technically do and in what kind of setting. Technical ability is not the same as robots showing up at work: the study estimates robots are cost competitive for only about 0.3% of tasks today.
Whenever Anthropic publishes new data. The current usage data comes from Claude conversations sampled in late 2025, and the robot ratings are from 2026.
Search covers 923 official occupations and tens of thousands of other titles people use. If yours does not show up, try a broader word like “manager” or “technician”, or browse the full list.
No. It is general information about occupations, not a prediction about your specific job, employer or future. Use it as one input among many.
An independent project tracking how AI changes work. It is not affiliated with Anthropic, OpenAI or the U.S. Department of Labor.
Free, no signup. Just type a job title.