Loading…
Loading…
How often is AI asked to do your job’s tasks? Rank among 718 jobs
Pick a job to see how often AI is asked to do its tasks
Built on the Anthropic Economic Index: millions of anonymised Claude conversations matched to the tasks of 718 occupations. Observed usage — not a job-risk score.
Share of all work conversations by job category, 2026-05 snapshot
Data: Anthropic Economic Index, 2026-05-01 period, retrieved 2026-09-05. Source and methodology. Numbers describe conversations matched to an occupation’s tasks — often from people who are not in that occupation. Snapshot only, no trend; it cannot show whether usage is rising, and it says nothing about job displacement or security.
Need productivity templates?
Task trackers, project planners, and automation scripts that save hours every week. Starting at $4.
Browse Productivity KitsEvery few weeks a headline claims AI will replace some profession. This tool replaces the headline with a number you can check: how often AI is actually asked to do the tasks of a given job, according to Anthropic’s published usage data. Search your title, see the rank, see whether the work is done together or delegated, and compare it with any other job.
The Anthropic Economic Index classifies anonymised conversations by the O*NET task they most resemble, then aggregates tasks up to occupations and to 22 job categories. Each occupation receives a usage share (percent of sampled work conversations), a rank among the 718 occupations with published data, and an augmentation/automation split of its own classified conversations.
This tool ships those rows as static data — no API call, no tracking — with an alias table that maps everyday titles (developer, CA, HR, teacher) to the closest O*NET occupation. Search suggests matches as you type; the head-to-head table places two occupations side by side.
The share card renders in your browser with the Canvas API. The numbers on it are the published figures, unchanged.
A copywriter checking whether "Writers and Authors" tasks are mostly augmentation (they are, at roughly 71%) before deciding how to pitch AI-assisted services.
A student choosing between data science and software development, comparing the two rows side by side.
A manager preparing an AI-adoption plan who wants to know which of the team’s roles have tasks that already show up heavily in AI conversations.
A journalist fact-checking a "top jobs using AI" listicle against the primary data.
Scope note: Occupations with broad task catalogues (document-management specialists, librarians) rank high because many different requests match their tasks, not because those professionals are the users. Only published rows are included; suppressed rows are missing, not zero. The data covers Claude usage, not every AI product, and is a single-period snapshot.
Type your job title — everyday titles like "developer", "nurse" or "CA" map to the closest published occupation.
Read the rank card: position among 718 occupations and the share of AI work conversations matching its tasks.
Check the augmentation vs automation bar to see whether people work with AI on these tasks or hand them off.
Add a second job to compare side by side, then share or download the card.
About the AI Usage by Job
The Anthropic Economic Index matches the content of anonymised Claude conversations to the task catalogue (O*NET) of each occupation. The share for a job is the percent of work conversations whose content matched one of that job’s tasks. It measures what AI is asked to do — not how many people in that job use AI, since the person asking is often not in the occupation.
The dataset cannot answer that, in either direction. It is a snapshot of observed usage with no trend series and no employment or income data, so it cannot show whether usage is rising and cannot support claims about displacement or security. Read the augmentation/automation split as a description of conversation styles, nothing more.
Augmentation conversations are collaborative — learning, iterating on a draft, validating an approach together. Automation conversations are directive — "do this task" with the person checking the output. Every occupation gets a split of its classified conversations between the two styles.
The index ranks 718 occupations, but publishes a row only where enough conversations exist to protect privacy. This tool carries the published rows for every job category plus the most-searched titles. If yours is missing, try a broader title or the nearest role — the tool suggests matches as you type.
From the public Anthropic Economic Index at anthropic.com/economic-index, data period 2026-05-01, retrieved 2026-09-05. Numbers describe Claude chat and Cowork usage only.
Spotify-Wrapped stats for ChatGPT, Claude & Claude Code
Open →UtilitiesYour whole life on one grid — one square per week
Open →UtilitiesFive rounds, one number — your reaction time in milliseconds
Open →UtilitiesGrab HD, SD & HQ thumbnails from any YouTube link
Open →