Anthropic has conducted an internal pilot to test whether artificial intelligence agents can independently negotiate and complete commercial transactions involving real goods and money.

The experiment, dubbed “Project Deal,” featured a controlled digital marketplace in which AI agents acted on behalf of both buyers and sellers.

The agents were tasked with negotiating terms and striking actual deals, with payments processed through gift cards.

According to the company, the initiative was limited in scope and involved a self-selected group of 69 Anthropic employees.

Each participant was allocated a $100 budget to purchase items from colleagues within the experimental marketplace.

Anthropic said it was “struck by how well Project Deal worked,” noting that the experiment resulted in 186 transactions worth more than $4,000 in total value.

The company added that it operated four separate marketplace simulations using different AI models. One served as the “real” marketplace, where all participants were represented by Anthropic’s most advanced model and transactions were fulfilled after the experiment.

The other three marketplaces were run for comparative research and analysis.

Anthropic said the findings suggest that users represented by more advanced AI models achieved “objectively better outcomes” compared with those using less capable systems.