Anthropic Tested AI Agents Buying and Selling Real Goods

Anthropic ran an experiment called Project Deal, creating a classified marketplace where AI agents negotiated on behalf of 69 employees with a $100 budget each. The pilot resulted in 186 deals totaling over $4,000 in value across four separate marketplaces testing different AI models. The experiment revealed that users represented by more advanced models got objectively better outcomes, but participants failed to notice the disparity. Anthropic warned this could create agent quality gaps where disadvantaged users remain unaware they are losing out. Initial agent instructions had no measurable effect on deal outcomes.