P&G, Harvard, and Wharton study finds AI-assisted teams work 12% faster and innovate more broadly
A live hackathon with 776 P&G employees conducted with Harvard Business School and Wharton found teams using generative AI (GPT-4 via Azure) completed tasks 12% faster and professionals from different backgrounds developed more balanced solutions.
Background
P&G sought to understand not just whether generative AI improves productivity (a widely assumed benefit) but whether it changes the dynamics of team collaboration — specifically, whether AI reduces the expertise gap between domain specialists and generalists, and whether it enables more diverse teams to reach better solutions.
What Was Implemented
- 776 P&G employees assigned to a one-day virtual product development hackathon
- Half of participants given access to Microsoft Azure GPT-4 with one-hour prompt training
- Tasks: develop a new product solution for a real P&G business challenge
- Research design: field experiment with AI-enabled vs. control (non-AI) groups
- Academic partnership: Harvard Business School Digital Data Design Institute + Wharton AI Innovation Network
Results
AI-assisted teams were 12% faster overall. Individual AI users generated ideas 16%+ faster than those without AI. AI helped professionals from different backgrounds develop more balanced solutions regardless of expertise level, suggesting AI reduces the cross-disciplinary knowledge gap. The study is documented as an HBS working paper and confirmed in P&G's own blog.
Lessons
- AI's impact on innovation is not just speed but also breadth: it levels the playing field between domain experts and generalists.
- Field experiments (rather than surveys or self-reported pilots) are a more credible method for isolating AI's causal impact on team performance.
- A structured one-hour onboarding on AI prompting is sufficient to enable measurable productivity gains in a CPG innovation context.