NYU Research Examines How AI Teammates Change Teamwork

Summary

An article about research by New York University professor J.P. Eggers and colleagues on how teams collaborate when AI tools become part of the group. In a six-hour buildathon, NYU students worked in small groups on problems faced by New York City residents, including grocery affordability, bike-lane safety and childcare access. They used AI for research, solution development and prototyping, while researchers observed their collaboration, interviewed participants and assessed the resulting solutions. The article identifies three main lessons. First, teams that worked in parallel through separate AI chats often created information overload, wasted effort and misalignment; teams should prompt together at important decision points and establish rules for individual versus shared AI use. Second, participants viewed AI as stronger at executing a well-defined direction than at discovering the right problem. Teams should therefore align on the problem first, then use AI more freely to prototype, visualize or code solutions. The article distinguishes “lean forward” AI use, in which people actively guide the system, from “lean back” use, in which AI performs more of the work. Third, the strongest teams were not necessarily the most technical. They combined different perspectives, including policy knowledge, personal experience and business thinking. The author argues that AI’s lower technical barrier makes it easier to involve outside experts such as lawyers, marketers and salespeople in collaborative prototyping. The conclusion is that AI produces more value when paired with aligned, diverse teams, while organizations may need to revisit team structures and norms around AI use.