Anthropic has connected its chatbot Claude to a new data pipeline. At the other end of this pipeline is Anthropic's own "economic index"—a vast database that records what people are actually doing with AI in the real world. Now, anyone using claude.ai can ask a question in the chatbox like they would ask a colleague, such as "Which professions use AI the most," and the answer will directly come from this index, rather than being made up by the model.

This feature, called the "Anthropic Economic Index Connector," was officially launched on July 23rd. It works with any model—whether you're using any version of Claude, in any conversation, you can activate it at any time, and no plugins or extensions are required. Open the connector menu on the left side of claude.ai, find "Anthropic Economic Index" in the directory, and turn it on. It takes about a minute for this "eye" to open.

This index itself serves as a benchmark for measuring how AI is actually used in the real economy. In the past, these numbers mainly served researchers, journalists, and policymakers; now, Anthropic clearly wants to put it in the hands of ordinary people—if you have even the slightest curiosity about whether "AI will take my department's livelihood," you can now explore the data yourself.

The usage is natural. You can start with broad questions, such as "What does this index say about my industry," and then dig deeper along the clues; or you can directly ask specific scenarios: What do people in Colorado use Claude for the most? What tasks do teachers usually ask AI to help with? What changes have occurred in the types of tasks automated by AI over the past year? Claude will provide answers and also show you the original data supporting those answers.

It's important to note that this mirror only reflects Claude's own usage patterns, not the entire labor market. Anthropic specifically reminded users during the release that the index reflects how people using Claude work, not the total picture of all AI tools. Therefore, before drawing conclusions from the data, Claude will guide you to review the source data and its limitations—what samples are not covered, and what conclusions should be questioned.