How to Use Claude’s Anthropic Economic Index Connector to Explore AI and Work

Introduction

Anthropic has added a new data connection to Claude that makes its economic research easier to explore through ordinary conversation.

The Anthropic Economic Index connector lets users ask questions about how Claude is being used across occupations, tasks, industries, countries, and U.S. states. Instead of answering only from general model knowledge, Claude can retrieve information from Anthropic’s published Economic Index data and show the evidence behind its response.

This makes questions such as the following much easier to investigate:

• Which occupations use AI most frequently?
• What tasks do teachers ask Claude to help with?
• How is Claude used differently across U.S. states?
• Are people delegating more complete tasks to AI over time?
• Which types of work are primarily augmented rather than automated?
• What does the available data suggest about a specific industry?

The connector became available on July 23,
2026. Anthropic describes it as a way for anyone to explore real data about AI and work directly inside Claude.

It requires no separate browser extension or local plugin. Once enabled, it can be used in normal Claude conversations and is not tied to one specific model.

The feature is useful, but its limitations are just as important as its convenience. The Economic Index measures patterns observed through Claude products and selected Anthropic API traffic. It is not a complete census of the labor market, all AI systems, or every worker’s experience.

Used carefully, the connector is a practical research window. Used carelessly, it can turn a narrow behavioral dataset into an overly broad claim about jobs and automation.

What Is the Anthropic Economic Index?

The Anthropic Economic Index is an ongoing research program designed to study how AI is entering real economic activity.

Anthropic analyzes privacy-protected patterns from Claude usage and publishes reports, interactive data, and downloadable datasets. The project has examined questions including:

• Which occupations and tasks appear most often in Claude conversations
• Whether AI is used to automate work or collaborate with a person
• How usage differs among countries and U.S. states
• How adoption relates to income and occupational structure
• What kinds of outputs people create with Claude
• How usage changes by time of day and day of the week
• How Claude Code, Claude chat, Cowork, and API use differ
• What users believe AI will mean for their jobs and careers

The Index is not one static spreadsheet.

It has developed through several data releases and research reports. Later editions introduced geographic comparisons, first-party API data, measures of task success and complexity, output classifications, higher-frequency telemetry, and survey results linked through privacy-preserving methods.

Anthropic also maintains an interactive Economic Index page and an open data repository on Hugging Face.

Before the connector launched, researchers typically had to read the reports, download data, inspect documentation, and perform their own analysis. Those options remain available.

The connector adds a natural-language interface on top of the published research.

What the New Connector Changes

The connector reduces the distance between a research question and the underlying data.

A user no longer needs to know the name of a specific release, find the correct CSV file, or write a data-analysis script before asking a basic question.

Claude can help with several stages:

1. Interpret the question.
2. Identify relevant Economic Index material.
3. Retrieve supporting data.
4. Explain the result in accessible language.
5. Show the source material behind the answer.
6. Continue into narrower follow-up questions.

For example, a broad opening question might be:

What does the Anthropic Economic Index say about AI use in education?

Claude can first provide an overview, then help the user narrow the investigation:

• Which educational occupations appear in the data?
• What tasks are teachers performing with Claude?
• Are those tasks categorized as augmentation or automation?
• How does educational use compare with software or management work?
• What limitations apply to the sample?

This conversational workflow is especially useful for people who do not work with statistical datasets every day.

Researchers can use it to identify relevant variables before downloading the raw files. Journalists can use it to test a story idea. Product teams can inspect whether a category of work appears in the data. Workers can explore how people in similar occupations are using Claude.

The connector does not replace the original data or methodology. It provides another route into them.

How to Enable the Anthropic Economic Index Connector

The exact labels in Claude may change as the interface is updated, but the setup is designed to be simple.

Step 1: Open Claude

Sign in to Claude through the web application.

The connector is an Anthropic-hosted integration, so there is no separate software package to install.

Step 2: Open the Connector Directory

Open the connector area from the Claude sidebar or navigate to the connector settings.

Depending on the current interface, this may appear under:

• Customize → Connectors
• Settings → Connectors
• The connectors section in the left sidebar
• The + menu beside the chat box

Step 3: Find Anthropic Economic Index

Search the connector directory for:

Anthropic Economic Index

Confirm that the listing is the official Anthropic connector.

Step 4: Connect or Enable It

Select the connector and follow the on-screen option to connect or enable it.

The Economic Index connector does not require a separate local installation. Unlike a connector for a private workplace service, it is accessing Anthropic’s public economic-research data rather than asking you to authenticate with an outside corporate account.

Step 5: Enable It in the Conversation

Claude connectors can be enabled or disabled for individual conversations.

If the connector is not already active, open the + menu near the chat box, choose Connectors, and turn on the Anthropic Economic Index connector.

Step 6: Ask a Grounded Question

Start with a clear question about an occupation, task, location, time period, output type, or form of AI use.

For example:

According to the Anthropic Economic Index, which occupations use Claude most often?
Please show the supporting source and explain the limitations of the comparison.

Asking Claude to show the relevant source and limitations is a useful habit. It reduces the chance that a concise answer will be mistaken for a complete labor-market conclusion.

Useful Questions to Ask

The connector works best when a broad question is gradually refined.

Occupation Questions

Which occupations appear most frequently in Claude work-related usage?

What kinds of tasks do teachers use Claude for?
Separate administrative, instructional, and content-creation work where possible.

How does Claude use in management occupations differ from use in computer and mathematical occupations?

Which occupations show the highest effective AI coverage, and how is that measure defined?

Task and Automation Questions

What share of observed Claude use is classified as automation rather than augmentation?
Has that changed across Economic Index releases?

Which work tasks are most often delegated directly to Claude?

Show examples of tasks where Claude collaborates with the user rather than completing the work autonomously.

What is the difference between task coverage and effective AI coverage?

Geographic Questions

What are the most common ways people in Colorado use Claude?

Which U.S. states use Claude more than expected relative to their working-age populations?

How does Claude use differ between high-adoption and low-adoption countries?

Compare AI usage patterns in Singapore, Australia, and the United States.
Explain how population adjustment changes the ranking.

Trend Questions

How has direct task automation changed across Anthropic Economic Index releases?

What changed in Claude Code usage between late 2025 and 2026?

How do work-related conversations differ between weekdays and weekends?

What types of outputs are becoming more common as Claude Code and Cowork grow?

Industry and Career Questions

What does the Index suggest about AI use in my industry?
My industry is professional marketing services.

Which parts of a teacher’s work appear most exposed to AI assistance?
Do not interpret exposure as job replacement.

What skills may become more important for people in occupations where Claude is used heavily?
Separate findings from the dataset from your own inference.

Does the Economic Index provide evidence that AI is eliminating jobs?
Explain what it can and cannot establish.

Research and Journalism Questions

Give me three defensible story ideas based on the Economic Index.
For each idea, identify the relevant dataset, the main limitation, and a claim that would be too strong.

Create a comparison table of Claude usage across selected occupations.
Include source release dates and definitions for each metric.

Which Economic Index findings have changed most across releases?
Link each statement to the relevant report or dataset.

What additional public data would be needed to test whether observed Claude usage is affecting employment?

A Better Prompt Structure for Reliable Answers

A useful Economic Index question usually contains four elements.

A stronger prompt might look like this:

Using the Anthropic Economic Index, compare Claude use in teaching occupations
and software-development occupations.

Include:
1. The most recent available data period
2. The main task categories
3. Whether usage is primarily augmentation or automation
4. The source for each major claim
5. Sampling and representativeness limitations
6. A clear warning against interpreting usage as job displacement

This format encourages Claude to behave more like a research assistant than a headline generator.

What the Data Can Tell You

The Economic Index can help answer descriptive questions about observed Claude usage.

It can show patterns such as:

• Which occupations are associated with more observed Claude tasks
• Which task categories are common
• How use differs by location
• Whether people tend to direct Claude or collaborate with it
• What outputs users produce
• How usage changes across products
• Whether usage patterns shift over time
• How users describe their experience of AI at work

The Index can also help researchers generate hypotheses.

For example, if one occupation shows unusually high direct automation, a researcher might investigate whether employers in that field are changing workflows, hiring practices, or role definitions.

That second investigation requires external evidence. The Index itself cannot complete it.

What the Data Cannot Prove

The connector makes data easier to access. It does not remove the need for careful interpretation.

It Does Not Represent Every AI Product

The Index primarily reflects activity involving Claude and selected Anthropic API traffic.

People may use ChatGPT, Gemini, Copilot, open-source models, industry-specific systems, local models, or internal enterprise tools differently.

A finding about Claude use is not automatically a finding about all AI use.

It Does Not Represent the Entire Workforce

Claude users are not a random sample of workers.

Knowledge workers, software professionals, managers, students, and people in countries with higher income and stronger internet infrastructure may be overrepresented.

Physical occupations and people with limited access to advanced AI tools may be underrepresented.

Anthropic’s own Economic Index Survey notes that its respondents are not representative of the general population.

Usage Is Not the Same as Job Replacement

Finding that Claude performs tasks connected to an occupation does not show that the occupation is disappearing.

A task may be:

• Partially automated
• Completed more quickly
• Performed with human review
• Expanded because the cost has fallen
• Shifted toward higher-level work
• Used only by a minority of workers
• Unsuitable for deployment in a high-stakes setting

Job effects depend on employer decisions, demand, regulation, wages, organizational change, complementary skills, and the creation of new work.

Exposure Is Not an Outcome

Measures such as task coverage or effective AI coverage describe the relationship between observed AI activity and occupational tasks.

They are not direct measurements of:

• Employment losses
• Wage changes
• Hiring reductions
• Productivity growth
• Hours saved across the whole economy
• Whether a company has adopted AI
• Whether a worker’s role has changed

Those outcomes require labor statistics, employer data, worker surveys, longitudinal studies, and causal research.

The Index Changes Over Time

Anthropic has repeatedly changed and expanded its methodology as Claude products have evolved.

Earlier usage centered on chat. Newer usage increasingly includes Claude Code, Cowork, tools, artifacts, and long-running agentic tasks.

Metrics from different releases may not always be directly comparable.

Claude should be asked to identify:

• The release date
• The sampled period
• The product surfaces included
• The population being measured
• The exact metric definition
• Any methodology change that affects comparison

Classification Is Not Perfect

Many Economic Index measures rely on automated classifiers.

A model may classify a conversation by occupation, task, artifact, purpose, or form of collaboration. Those classifications can be useful at scale, but they are estimates rather than perfect labels.

Ambiguous conversations, multilingual content, unusual jobs, and mixed-purpose sessions may be difficult to categorize.

Privacy and Data Handling

Anthropic describes the Economic Index as using privacy-preserving analysis.

In the June 2026 report, Anthropic states that its analyses use privacy-preserving classifiers and suppress cells with insufficient observations. The report says transcripts used in the research pipeline are read by another Claude instance rather than by human researchers during the classification process.

Published datasets contain aggregated results rather than a public archive of identifiable user conversations.

The connector exposes those published Economic Index resources. It is not presented as a tool for searching an individual person’s private Claude history.

Users should still distinguish between two separate things:

1. The public Economic Index connector, which queries published aggregate research data
2. Private connectors, which may connect Claude to a user’s workplace tools or accounts

A public research connector does not grant access to another person’s conversations.

How to Verify Claude’s Answer

A polished answer can still contain a mistaken interpretation.

Use the following verification process for important research.

1. Ask for the Exact Source

Request the report title, release date, dataset, and relevant table or chart.

Which report and data release support this claim?
Give me the exact source title and publication date.

1. Ask for the Metric Definition

Terms such as automation, augmentation, task coverage, effective AI coverage, usage index, and success rate have specific definitions.

Define this metric using Anthropic’s methodology.
Do not replace the technical definition with a general explanation.

1. Check the Denominator

A percentage is difficult to interpret without knowing what was counted.

Ask whether it represents:

• Conversations
• Tasks
• Users
• Occupations
• Time-weighted duties
• Survey respondents
• API requests
• A normalized geographic index

1. Check the Sample Period

AI usage can change quickly.

A finding from early 2025 may not describe usage in mid-2026. Ask Claude to state the observation window rather than only the article’s publication date.

1. Open the Original Report

Use the connector to discover the evidence, then inspect the primary page or downloaded data when the claim matters.

1. Compare with External Data

For labor-market conclusions, compare the Index with sources such as:

• Government employment statistics
• Occupational databases
• Wage and vacancy data
• Employer surveys
• Worker surveys
• Productivity studies
• Industry-specific adoption research

The Economic Index is one evidence stream, not the entire economic picture.

Example Research Workflow

Suppose a product team wants to understand AI use among teachers.

Step 1: Begin Broadly

What does the latest Anthropic Economic Index say about Claude use in teaching occupations?

Step 2: Identify the Data

Which report, release, occupation codes, and observation period support that answer?

Step 3: Separate Tasks

Group the observed teaching-related tasks into lesson preparation,
administration, assessment, communication, research, and direct student support.

Step 4: Distinguish Automation from Collaboration

Which task categories appear more automated, and which involve iterative collaboration?
Explain the classification method.

Step 5: Identify Missing Evidence

What can this dataset not tell us about classroom adoption,
student outcomes, teacher workload, or employment?

Step 6: Download the Data

Use Anthropic’s Economic Index repository when a reproducible table, chart, or statistical analysis is required.

Step 7: Validate with Real Users

Interview teachers, compare school policies, inspect local adoption data, and test whether the Index’s categories match actual workflows.

The connector can accelerate the first half of the research process. It cannot replace fieldwork.

Who Is the Connector Useful For?

Researchers

Researchers can use it to locate relevant releases, clarify metric definitions, compare patterns, and generate hypotheses before working directly with the raw data.

Journalists

Journalists can quickly test whether a claim about AI and work is supported by the Index and identify the limitations that should appear in the article.

Policymakers

Policy teams can explore geographic and occupational patterns, then connect them to official labor statistics and stakeholder interviews.

Product Teams

Teams building workplace AI products can investigate which tasks already appear in Claude usage and where user behavior may be changing.

Employers

Employers can use the Index as a broad reference while avoiding the mistake of assuming that aggregate Claude usage describes their own workforce.

Workers and Students

Individuals can explore how similar occupations use Claude, identify emerging workflows, and generate questions for further career research.

The connector should not be used as an automatic career forecast. It is better suited to understanding observed task patterns than predicting the fate of one person’s job.

Why Anthropic Is Putting the Index Inside Claude

The Economic Index was originally most accessible to researchers, economists, journalists, and policy specialists who were willing to read long reports or analyze datasets.

The connector broadens the audience.

A person curious about AI’s effect on their department can now ask a direct question without first learning the structure of Anthropic’s data releases.

This also changes the role of Claude.

The model is not only the subject of the Economic Index. It becomes the interface through which people examine the data created from Claude usage.

That creates a useful feedback loop, but it also requires transparency. Claude should make it clear when an answer comes from the Economic Index, which release it used, and how narrow the evidence is.

A conversational interface can make complex data more understandable. It can also make uncertain findings sound more definitive than they are.

The strongest use of the connector is therefore not asking Claude to deliver one final verdict on the future of work.

It is using Claude to:

• Form better questions
• Locate the relevant evidence
• Compare definitions
• Expose uncertainty
• Inspect source data
• Identify what additional research is needed

常见问题

What is the Anthropic Economic Index connector?

It is an official Claude connector that lets users explore Anthropic’s published data about how Claude is used across tasks, occupations, products, and locations. Claude can answer questions using the Index and point users back to supporting sources.

Do I need to install a plugin?

No separate browser extension or local plugin is required. The connector can be enabled from Claude’s connector directory and activated for a conversation.

Does the connector work with every Claude model?

Anthropic’s launch information says the connector can be used across Claude models and conversations. The interface and available model choices may change over time.

Is the Anthropic Economic Index real labor-market data?

It is real observational data derived from Claude usage and related Anthropic research, but it is not a complete labor-market census. It should be combined with employment statistics, surveys, and other evidence before making broad claims.

Can the connector tell me whether my job will be replaced?

No. It can show how tasks related to an occupation appear in Claude usage, but it cannot reliably predict one worker’s job outcome or prove that observed AI use will cause employment loss.

Does the Index include data from ChatGPT or Gemini?

No. The Economic Index is based on Claude-related data and selected Anthropic API activity. Its findings should not be generalized automatically to every AI platform.

Is the raw Economic Index dataset publicly available?

Yes. Anthropic publishes Economic Index data releases through its repository on Hugging Face, with documentation for the different releases.

How can I make Claude cite the data correctly?

Ask for the exact report, data-release date, observation period, metric definition, denominator, and relevant table or chart. For important work, open the primary source and verify the result directly.

相关工具

• Claude: Anthropic’s conversational AI interface where the Economic Index connector can be enabled.
• Claude Connectors: The official directory for integrations that connect Claude with external tools and data.
• Anthropic Economic Index: The interactive public portal for exploring Anthropic’s economic research.
• Anthropic Economic Index Dataset: The official repository containing documented Economic Index data releases.
• O\NET OnLine: A U.S. occupational database useful for interpreting task and occupation mappings in labor research.

Related Links

• Anthropic Economic Index Connector Announcement: Anthropic’s official introduction to querying Economic Index data inside Claude.
• Anthropic Economic Index Home: The interactive index and entry point to the research program.
• Economic Index Report: Cadences: The June 2026 report covering usage rhythms, outputs, surveys, and updated methodology.
• Economic Index Geography Report: Research on Claude adoption across countries, U.S. states, and business API use.
• Labor Market Impacts of AI: Anthropic’s framework for comparing observed AI use with occupational task exposure.
• Economic Index Data Repository: Downloadable releases, documentation, licensing, and citations.
• Claude Connector Setup Guide: Official guidance on connector configuration and per-conversation enablement.

Summary

The Anthropic Economic Index connector brings a public economic-research dataset into ordinary Claude conversations. Users can ask about occupations, tasks, regions, automation, collaboration, and changes in how Claude is used without beginning with manual data analysis.

Its greatest value is accessibility. Researchers, journalists, policymakers, product teams, employers, and workers can use natural language to locate evidence and refine questions.

The connector does not turn Claude usage into a complete picture of the economy. Its data reflects a non-representative subset of AI activity, uses automated classifications, changes across releases, and cannot by itself prove job displacement or other causal labor-market effects.

Use the connector as a guided route into the evidence—not as an oracle that can deliver a final answer about the future of work.