Most business owners deciding where to start with AI are guessing. You read a headline, you hear a pitch from a software vendor, you notice a competitor bragging about something on LinkedIn. What has been missing is a plain answer to a plain question: what are people in my line of work actually handing to AI right now?
On July 22, 2026, Anthropic — the company behind the Claude assistant — announced that it had put the data behind that question into a connector you switch on inside Claude and then talk to. Anthropic says "getting started takes about a minute," and that there is "nothing to install."
What Anthropic actually released
Every fact and quotation below comes from Anthropic's own announcement, published July 22, 2026. It is the only source for this story, and where Anthropic states a limitation, we have carried it over rather than trimmed it.
Anthropic maintains something called the Anthropic Economic Index. In Anthropic's words, the Index "measures how AI is actually being used in the economy" — not what people say in a survey that they plan to use it for, but the patterns that show up in real usage.
That data already existed. The problem was who could reach it. Anthropic writes that the Index "has been useful to researchers, journalists, and policymakers, but we want it to be just as accessible to anyone curious about how AI fits into their field or day-to-day life." In other words, it lived as datasets and research reports — a fine format for an economist, a poor one for someone running a plumbing company or a 40-person accounting firm.
The connector changes the doorway, not the data. Instead of downloading a spreadsheet and knowing what to do with it, you turn on a feature inside claude.ai and ask questions in ordinary English. Anthropic's phrasing for what comes back: "You'll get answers grounded directly in the Index data."
That is the entire product. It is not a new AI model and it is not a business tool that does work for you. It is a way to ask a research dataset a question out loud.
Why this matters if you run a business
The hardest part of adopting AI is not the technology. It is picking the first thing to point it at. Choose badly and you burn a month of somebody's time, the pilot quietly dies, and the organization concludes that AI does not work here. That conclusion then sticks for a year.
Here is what a tool like this changes about that decision:
- It replaces a hunch with a starting point. "Where do people in my industry already get value from this?" is a better first question than "what did the software rep tell me last Tuesday?"
- It is about tasks, not job titles. Anthropic's examples are phrased around what kinds of work get handed over — which is the level a business owner can act on, because you buy and assign tasks, not occupations.
- It shows its work. Anthropic explicitly tells you to "ask Claude to show you the underlying data behind any answer." Most AI answers cannot be traced anywhere. This one is designed to be.
- The cost of looking is close to zero. A minute of setup and a few questions is cheaper than a consultant's first meeting, and cheaper than the wrong pilot.
None of that guarantees you find something useful. Anthropic does not claim the Index will improve any business outcome, and neither do we. The claim is narrower and more honest: it beats guessing where to start.
How to turn it on
Anthropic's instructions are one sentence long: "In claude.ai, open the connectors menu, find the Anthropic Economic Index in the directory, and enable it." Anthropic adds that it "works in any conversation with any Claude model," and again, there is nothing to install.
- Sign in to claude.ai in a web browser.
- Open the connectors menu.
- Find the Anthropic Economic Index in the directory.
- Enable it.
- Start a normal conversation and ask your question.
A note on price, because this is where articles usually start inventing things. Anthropic's post says "the full datasets remain freely available on our website," and it describes the connector as something you find in a directory and switch on with nothing to install. The post does not name a price for the connector itself, and we are not going to make one up. The reasonable expectation is that this is a setting inside an account you may already have, not a purchase — but check the connectors directory in your own account before you put a number in a budget.
What to ask it
Anthropic's advice is to start broad and then drill in. Its suggested opener is simply: "What does the Index say about my industry?"
These are the example questions Anthropic published:
- "Which occupations use AI the most?"
- "What are the most common ways people in Colorado use Claude?"
- "What sorts of tasks do teachers use Claude for?"
- "What kinds of tasks are people automating with AI? How has that changed over the past year?"
Swap in your own field and your own state and you have your first three questions. Then follow Anthropic's other instruction every single time: ask it to show you the underlying data behind the answer. An answer you cannot trace back to a number is a rumor with better grammar, and that rule applies to every AI tool you will ever use, not just this one.
The limitation Anthropic states itself
This is the part most coverage will skip, so read it carefully. Anthropic writes: "the Index reflects patterns in Claude usage rather than the labor market as a whole."
Translated: this is a record of what people did inside one company's AI assistant. It is not a census of the economy. Think of it as a very detailed survey of everyone who walked into one particular store — genuinely informative about that store's customers, and silent about everyone who shopped somewhere else or did not shop at all.
Three practical consequences for a business owner reading the results:
- Early-adopting fields will look busier than they are. Industries full of people who reached for an AI assistant first will be over-represented. That is a fact about who signed up, not about who benefits.
- Work locked inside your own systems is largely invisible. Tasks that live in an ERP, a CRM, a scheduling system or a regulated records system rarely get pasted into a chat window, so they show up thinly here — even where AI could help most.
- "Nobody uses AI for this" may just mean "nobody uses Claude for this." An absence in the data is not proof that something does not work. It can equally be an opening.
Anthropic also says that "Claude will point you back to the source data and its limitations as you explore." Take it up on that. A vendor that ships the caveat alongside the product is giving you something you can actually use.
How to use it without fooling yourself
An hour with this, done properly, is worth more than a week of reading vendor blogs:
- Write your own list first. Before you open it, list the five tasks in your business that eat the most hours and produce the least judgment — the retyping, the summarizing, the chasing, the formatting. That list is yours and it is real.
- Then compare. Ask what the Index says about your industry and see where the two lists overlap. Overlap is where you start, because someone else has already proven the task is a reasonable fit.
- Ask for the data behind every claim, and read it. If it will not show you, do not repeat it.
- Do not present it as market research. It is one vendor's usage data. It is a compass, not a survey of your customers, and it should never appear in a board deck as if it were.
- Come back to it. Anthropic's own example question asks how things changed over the past year, which means the picture moves. What was absent last quarter may not be absent next quarter.
The honest summary
A research dataset got a conversational front door. That is a smaller story than the headlines around AI usually promise, and a more useful one than most, because the thing standing between a lot of businesses and their first sensible AI project is not capability — it is not knowing where to point.
It will not tell you what to do. It will tell you what other people in your field have already tried, with the caveat that "other people in your field" means the ones who used Claude. For a one-minute setup, that is a reasonable trade. Spend the minute, ask the three questions, ask for the data behind the answers, and treat whatever you find as a starting point rather than a verdict.
We break down AI stories like this one for business owners at the AIWAS Knowledge Base, and you can follow the short versions on @AIWASai.