The Cybernetic Teammate: A Field Experiment on Generative AI and Teamwork
AI matched the performance of human teams without AI in a P&G experiment and helped bridge R&D-commercial silos.
Summary
In a randomised experiment with several hundred P&G professionals working on real product-innovation challenges across four business units, teams paired a commercial expert with an R&D specialist, and some participants worked alone, with and without GPT-4 access. Teams without AI outperformed individuals without AI, and individuals using AI matched the performance of teams without it, while AI-enabled teams were best and produced top-decile solutions significantly more often. AI users saved 12 to 16% of their time and wrote more detailed solutions. Without AI, R&D specialists proposed technical solutions and commercial experts market-led ones; with AI, both produced balanced, integrated solutions, and less experienced staff performed like more experienced colleagues. AI users also reported more positive and less negative emotion. The authors argue that AI acts as a teammate, not just a tool, with consequences for team design, training and the boundaries between functions.
Editor's opinion
Generative AI can stand in for part of a team: an individual with AI matched a team without it, and the technical-versus-commercial divide between specialists narrowed. Companies should rethink how they staff innovation work, since AI lets fewer people cover more of the functional perspectives, and should watch for the risk that expertise is flattened along with the silos.
Key concepts
generative AI · teamwork · new product development · field experiment · Procter & Gamble