This guide shows how to check an occupation against cited task-use data. The setup follows official documentation checked 26 October 2026. BIG CHANGE did not run the connector in a Claude account.

The big change

  • What changed: The connector brings occupation data into Claude.
  • Why it matters: Match listed tasks to your duties and open their citations.
  • What to watch: Check the data period and denominator. Usage share cannot predict job loss.

Before you start

You need an authorized Claude account and permission to use the connector. Write down your occupation as you call it, plus several duties you actually perform. Use task descriptions you may share; leave out employer secrets, personal records and details about coworkers. Check your account's usage allowance and enabled usage credits before a long session. Connector calls consume plan resources; there is no fixed price for this exercise in the sources reviewed here.

O*NET labels can differ from job titles. Check the listed tasks. If no occupation reasonably fits, record that and stop short of a conclusion about your own work.

1. Find and enable the connector

In claude.ai, open Customize > Connectors, then use the plus control to browse the directory, as Claude's help says. Select Anthropic Economic Index if listed. Review its capabilities and use the setup action shown, such as Connect or Install. We have not observed the Index-specific dialog. For Team or Enterprise, an Owner or Primary Owner may need to enable the connector for the organization.

In the conversation, open the + menu, choose Connectors, and enable the desired service for that conversation. Confirm Anthropic Economic Index is selected before sending the prompt. If it is missing or permission is denied, check account and organization settings. Mark the result incomplete until access is resolved.

2. Ask for the closest occupation and its records

Fill the blanks with your own permitted details, then send the prompt. It supplies no sample answer.

Prompt
My job title: <your title>
My actual duties:
<list permitted duties>

Find the closest occupation
in the Economic Index by
task match. Give its exact
occupation label and say
which of my duties fit or
do not fit. Do not infer a
personal job risk.

List that occupation's
Claude tasks. Link records.
Name dataset, period,
population, product, measure
and denominator. Explain
automation and augmentation.
Mark missing fields.

Expect an occupation match and linked task records. These are requests for the reader to try, not results BIG CHANGE obtained. If Claude gives several plausible occupations, compare their task lists and keep the ambiguity in your record.

3. Check the answer at its source

Check cited occupation and task descriptions against your duties. Record the period in the cited data; the Index page update date may differ. Before repeating a percentage, identify its denominator. Claude-use, task and workforce shares cover different populations.

An unobserved task proves neither capability nor job safety. If records are inaccessible, the period is missing, or a percentage has no traceable denominator, leave that figure out of the conclusion and mark the check incomplete.

Keep a private result record. Write “missing” where evidence is unavailable:

Snippet
My title: ____
My duties: ____
Index occupation: ____
Task match/gaps: ____
Dataset: ____
Data period: ____
Sample/population: ____
Product covered: ____
Measure: ____
Denominator: ____
Task data: ____
Automation labels: ____
Source URLs: ____
Limits/conflicts: ____
Conclusion: ____
Status: complete/incomplete

4. Say only what the records support

Index data covers Claude use; labels classify automation and augmentation. Anthropic's separate observed-exposure research combines task capability and work-related use, weighting automated use more heavily; it is a different research measure. None of these is a personal probability of job loss. They also leave other AI products, task success, changes in job design and your local labor market outside this check.

Write one sentence naming the matched occupation, data period, product, measure and cited task pattern. Add one sentence naming the task mismatch or missing evidence. If the occupation fit, records, period or denominator fails inspection, use incomplete in the record and explain why.

For broader context on why vendor use data is hard to generalize across work, see BIG CHANGE's earlier analysis of AI use beyond coding.