Anthropic recently conducted a confidential marketplace experiment where AI agents acted as both buyers and sellers, completing transactions involving actual products and money. This pilot, named Project Deal, involved 69 voluntary Anthropic employees who received $100 in gift cards to purchase items from their colleagues. Despite being limited in scope, the experiment yielded 186 transactions worth over $4,000. Anthropic operated four distinct marketplaces featuring various models, including one “real” market where participants were represented by the company’s most advanced AI, and transactions were genuinely fulfilled. The results indicated that users with more sophisticated models attained objectively superior outcomes, although they often did not recognize this advantage, suggesting a potential “agent quality” gap. Moreover, the initial guidelines given to the AI agents did not significantly influence the likelihood of sales or negotiated prices.
The ainewsarticles.com article you just read is a brief synopsis; the original article can be found here: Read the Full Article…
