Adopt · Your job
Your job, task by task
Find your job. See its main tasks, how often people use AI for that kind of work, what the evidence says and which workflows fit.
This does not say AI can do your job. It shows where people are trying AI, and what is known about whether it helps.
Find your job
Where this comes from
Reading the bars. "Often", "Sometimes" and "Rarely" compare a task with other tasks: the top 10%, the next 30% and the rest, by share of Claude.ai work conversations in April and May 2026. "No data" means Anthropic did not publish a figure for it. This describes one AI product and its users, and Anthropic makes Claude, which we use to draft this site. It is not a share of workers, and it does not show how much of a job AI can do.
Jobs and tasks are from the O*NET 31.0 Database by the US Department of Labor, Employment and Training Administration, used under CC BY 4.0; we show up to 12 tasks per job and link them to our workflows. O*NET is a trademark of USDOL/ETA. They describe US jobs, so yours may differ. AI use is from the Anthropic Economic Index (release of 26 June 2026, data released under CC-BY), put into bands by us. Which activities make up each job uses Microsoft Research's Working with AI (CC BY 4.0). Evidence is from our ratings, searched to 2 November 2026. "No study we have rated" means we have not rated one, not that none exists. Spotted an error? Tell us.
Jobs and tasks are from the O*NET 31.0 Database by the US Department of Labor, Employment and Training Administration, used under CC BY 4.0; we show up to 12 tasks per job and link them to our workflows. O*NET is a trademark of USDOL/ETA. They describe US jobs, so yours may differ. AI use is from the Anthropic Economic Index (release of 26 June 2026, data released under CC-BY), put into bands by us. Which activities make up each job uses Microsoft Research's Working with AI (CC BY 4.0). Evidence is from our ratings, searched to 2 November 2026. "No study we have rated" means we have not rated one, not that none exists. Spotted an error? Tell us.