According to AnthropicAI on Twitter, Anthropic launched Project Deal, a controlled marketplace inside its San Francisco office where Claude handled buying, selling, and negotiation on behalf of employees, executing end‑to‑end dealmaking tasks (source: Anthropic on X, April 24, 2026). As reported by Anthropic, the experiment evaluates Claude’s agentic capabilities in price discovery, counteroffers, and closing, highlighting practical applications for autonomous procurement, internal resale programs, and B2B negotiation workflows (source: Anthropic on X). According to Anthropic, the setup used real participants and real items, enabling measurement of negotiation success and user satisfaction—key metrics for deploying AI negotiators in enterprise marketplaces and expense management (source: Anthropic on X).

According to Anthropic on X, the team used Claude to interview 69 employees about items they wanted to buy and sell, collected custom instructions for each participant, and then deployed agents to haggle across four parallel markets to test how different model variants negotiated. As reported by Anthropic, the controlled setup isolates the impact of model choice on price discovery, concession patterns, and deal completion rates, offering a practical benchmark for multi-agent negotiation performance. According to Anthropic, this design enables businesses to evaluate which Claude versions yield better transaction outcomes, faster time to deal, and improved surplus capture in marketplace, procurement, and sales automation use cases.

According to AnthropicAI on Twitter, simulated negotiations between Claude Opus and Claude Haiku agents showed Opus consistently securing substantially better deals, while human survey participants failed to perceive the gap, as reported by Anthropic’s post and study snippet. According to Anthropic, the result underscores how higher‑capability LLMs can translate model quality into tangible economic outcomes in automated bargaining and procurement workflows. As reported by Anthropic, this perception gap creates operational risks for enterprises that evaluate agent performance by intuition rather than outcome metrics, suggesting demand for rigorous A/B testing, revealable logs, and controllable negotiation policies in agentic systems. According to Anthropic, organizations deploying multi‑agent systems for sourcing, ad bidding, or dynamic pricing can realize measurable ROI by upgrading from lighter models to stronger models like Opus where negotiation or strategic reasoning is core.