Generative AI can effectively function as a "cybernetic teammate" by replicating key benefits of human collaboration, including performance enhancement, expertise sharing, and positive emotional engagement.

Objective: This study aimed to examine how generative AI transforms the core pillars of collaboration—performance, expertise sharing, and social engagement—in workplace team settings, moving beyond viewing AI as merely a tool to exploring its potential as an active participant in collaborative processes.

Methods: The researchers conducted a pre-registered field experiment with 776 professionals at Procter & Gamble (P&G), a global consumer packaged goods company. Participants worked on real product innovation challenges across four business units, randomly assigned to one of four conditions in a 2x2 experimental design: (1) individuals working without AI, (2) teams of two (one R&D and one Commercial professional) working without AI, (3) individuals working with AI, and (4) teams working with AI. The experiment followed P&G's standard new product development process, with participants developing solutions for authentic business challenges in their units. Performance was measured through expert evaluations of solution quality, while expertise sharing was assessed by analyzing the balance of technical versus commercial content in solutions. Emotional responses were collected through pre- and post-task self-reported measures.

Key Findings:

• AI significantly enhanced performance: individuals working with AI produced solutions of comparable quality to teams working without AI, demonstrating that AI can effectively substitute for certain collaborative functions.
• Teams with AI and individuals with AI both showed substantial quality improvements (0.39 and 0.37 standard deviations, respectively) over the baseline of individuals working alone without AI.
• AI broke down functional silos: Without AI, R&D professionals tended to suggest more technical solutions, while Commercial professionals favored commercially-oriented proposals. When using AI, professionals generated more balanced solutions regardless of their background.
• AI provided positive emotional experiences: Participants using AI reported higher levels of positive emotions (enthusiasm, energy, excitement) and fewer negative emotions (anxiety, frustration) compared to working alone without AI, matching emotional benefits typically associated with human teamwork.
• AI-enabled teams were three times more likely to produce solutions ranking in the top quality decile, highlighting AI's impact on exceptional performance.
• Participants using AI completed tasks in significantly less time (16.4% less for individuals, 12.7% less for teams) while producing substantially longer and more comprehensive solutions.
• AI particularly benefited employees less familiar with product development tasks, allowing them to achieve performance levels comparable to teams with at least one experienced member.

Implications: The findings suggest that AI is no longer merely a passive tool but functions as a "cybernetic teammate" that can replicate key benefits of human collaboration. By demonstrating that AI-enabled individuals can match team-level performance while breaking down expertise boundaries, the study challenges traditional assumptions about team structures and organizational design. Organizations may need to fundamentally rethink optimal team compositions, work processes, and skill development approaches as AI adoption scales, with potential for more flexible and efficient organizational structures.

Limitations: The study has several limitations. First, although it followed the firm's early-stage product development routine, the experiment relied on one-day virtual collaborations that may not fully capture the complexities of long-term team interactions, such as extended coordination challenges and iterative workflows. Second, the focus on cross-functional pairs may not represent how AI integration would function in larger, more complex team structures or with teams of similar expertise. Third, participants were relatively inexperienced with AI prompting techniques, suggesting the observed benefits may represent a lower bound of potential advantages as users develop more sophisticated AI interaction strategies.

Future Directions: The authors suggest several avenues for future research: (1) examining how the benefits of AI integration evolve as users become more sophisticated in their AI interactions, (2) investigating what specific features of AI systems support effective knowledge integration across professional boundaries, (3) exploring how organizations can effectively capture and disseminate best practices for AI-enabled work, and (4) studying how AI integration affects the development of domain expertise over time—whether it leads to genuine expertise development or primarily facilitates access to existing knowledge.

Title and Authors: "The Cybernetic Teammate: A Field Experiment on Generative AI Reshaping Teamwork and Expertise" by Fabrizio Dell'Acqua, Charles Ayoubi, Hila Lifshitz, Raffaella Sadun, Ethan Mollick, Lilach Mollick, Yi Han, Jeff Goldman, Hari Nair, Stew Taub, and Karim R. Lakhani.

Published On: March 21, 2025

Published By: Harvard Business School Working Paper 25-043