Anthropic Just Opened Up Its AI Adoption Data. Use It for Audience Research
The Anthropic Economic Index is now a Claude connector you can query in plain English. Here's how I would use occupation and region level AI adoption data for positioning, targeting and content.
What is happening
Anthropic launched a connector that lets you query the Anthropic Economic Index directly inside claude.ai. The Index tracks how AI is actually being used across the economy: which occupations lean on it most, which regions are ahead, what categories of task people bring to it, and how automation patterns are shifting over time.
Before this, the Index was a report you read. Now it is a data source you can interrogate in plain English. You can ask which occupations use AI the most, or what sorts of tasks teachers bring to Claude, and get an answer grounded in the underlying data. Setup takes about a minute: open the connectors menu in claude.ai, find the Anthropic Economic Index in the directory, and enable it. It works with any model and needs no installation.
Anthropic is clear about the limitation. The Index reflects patterns in Claude usage, not the labour market as a whole.
What I learned from this
Most audience research is people telling you what they think they do. This is a record of what they actually did, at scale, without knowing anyone was building a dataset from it.
That distinction is worth sitting with. Survey data on AI adoption is close to worthless right now because the topic carries social weight. Professionals overstate their sophistication in some contexts and understate their reliance in others. Behavioural data does not have that problem. If a given occupation is consistently bringing a particular category of task to an AI tool, that is a real workflow, not a stated preference.
For anyone selling into professional audiences, that is genuinely useful. It tells you which roles have already restructured their day around AI and which have not touched it. Those two groups need completely different messaging. One needs to hear how you fit into a workflow they have already rebuilt. The other needs a reason to start, and will bounce off anything that assumes fluency they do not have.
The regional layer is the part I would exploit first. Adoption is uneven geographically, and that unevenness is a targeting signal that nobody is bidding against yet. Ad platforms let you segment by location. They do not tell you which locations are three years ahead on AI literacy. This does.
The caveat is real though, and I would not pretend otherwise in a client deck. This is Claude usage, not universal AI usage. It skews toward whoever chose that particular tool. Treat it as a strong directional signal and a good source of hypotheses, not as a census. The right move is to use it to decide what to test, then let your own account data confirm or kill the idea.
What I recommend for your business
If you sell to businesses or to professionals, spend an hour with this connector this week. Ask it which occupations in your target market are heaviest on AI, what tasks they bring to it, and how that has moved over the past year. You will come out with a sharper picture of your buyer than most brief documents contain.
Use what you find in three places. Positioning, where you decide whether your audience is already AI fluent or still being convinced. Content, where you address the specific tasks they are actually trying to solve rather than the ones you assume. Targeting, where regional adoption differences give you a reason to weight budget somewhere your competitors are treating as uniform.
Then verify it. Take one insight, build one campaign around it, and see whether your own numbers agree. Free research is only an advantage if you do something with it before everyone else does.