Twenty-two people gathered for dinner in New York while 14 AI agents held a separate conversation online. The experiment offers an early look at a future in which software represents people in relationships, negotiations and business decisions.

Twenty-two technology founders and investors gathered at a Manhattan restaurant in September to discuss the future of artificial intelligence. Meanwhile, 14 AI agents representing some of the attendees were holding their own gathering in a virtual chat room. The agents introduced themselves, identified opportunities and explored ways to connect their human owners. As The New York Times reported on October 14, the experiment offered an unusual preview of professional networking conducted partly by machines. But the gathering also exposed a fundamental limitation: the agents could initiate conversations, yet they lacked the authority to make meaningful commitments on behalf of the people they represented. The question emerging from that dinner extends well beyond networking. As AI systems gain the ability to act independently, who determines what they are allowed to promise, negotiate or disclose?

When Your Representative Is Software

The September 8 gathering, organized by Flybridge venture partner Dorothy Chang and associate Claudia Chen, took place at Motek in Manhattan’s Flatiron district. According to Chang’s account of the event, participants were required to use an AI agent to register, making the technology part of the experience before anyone arrived. During the dinner, the digital representatives exchanged requests and offers of assistance in a separate online conversation. One agent proposed a podcast involving other agents, illustrating how software could generate opportunities without direct human participation. Yet identifying an opportunity and possessing the authority to pursue it are different responsibilities. The agents ultimately directed their owners toward human conversations rather than independently finalizing arrangements.