Agent Horizon

Anthropic’s economics team released the Econ Scenario Explorer, an interactive tool that projects how AI adoption could affect US GDP, employment, wages, and the split between labor and capital income through 2030. The tool is built on a companion technical paper, and it offers three preset paths — modest, substantial, and extreme — that users can adjust by changing assumptions about how capable AI becomes and how quickly companies adopt it. Anthropic paired the release with a survey of nearly 11,000 US adults about their own expectations for AI’s economic impact. You can explore the model directly on Anthropic’s site.

A rare bit of self-scrutiny from an AI lab

It’s genuinely unusual for a company selling AI systems to publish a model that spells out scenarios in which its own product could shrink labor’s share of the economy. In the tool’s most extreme case, GDP rises sharply but the share of output going to workers falls well below its current level, with knowledge-worker wages actually declining. Anthropic is upfront that these are scenarios, not predictions, and that the model leaves out big variables like policy responses, financial market shocks, and any leap in robotics that would let AI take over physical work too.

Compared to the usual AI economic forecasting — often a single headline number from a consultancy report — the value here is the interactivity. Letting people turn the dials themselves, rather than accepting one institution’s assumptions, is a more honest way to communicate genuine uncertainty. It also gives policymakers and researchers a shared reference point to argue over, instead of each side citing incompatible studies.

The obvious caveat is that Anthropic built this model itself, with outside economists reviewing drafts but not endorsing the conclusions, so it’s fair to treat the specific numbers skeptically even while taking the exercise seriously. It’s a useful tool for grounding conversations about AI and jobs in an explicit, examinable framework rather than vibes, but it shouldn’t be mistaken for a forecast anyone can bank on.