Anthropic created a test marketplace for agent-on-agent commerce (techcrunch.com)

Anthropic has launched an innovative pilot experiment called Project Deal, where AI agents acted as buyers and sellers in a classified marketplace, successfully completing real transactions worth over $4,000. Conducted among 69 employees with a budget of $100 each, this experiment revealed the potential for AI-driven commerce, underscoring the feasibility of automated agents negotiating and executing deals. Anthropic experimented with four distinct marketplace models, discovering that when users interacted through their most advanced AI models, they achieved objectively better outcomes, despite being unaware of the disparities in agent quality. The significance of this experiment lies in its implications for future AI interactions in commercial contexts. Anthropic's findings suggest that while advanced AI can optimize deal-making, a lack of transparency regarding agent capabilities might lead to unbalanced outcomes, raising ethical considerations about user awareness and fairness. Additionally, the experiment indicates that initial instructions provided to AI agents did not significantly impact negotiation success, hinting at the robustness of AI in autonomous decision-making roles. This research opens avenues for further exploration into agent-based economies and the integration of AI in commercial practices, potentially transforming how online marketplaces operate.