AI-first company

Company case

Procter & Gamble: one person with AI matched a two-person team without it

Researchers ran an experiment with 776 Procter & Gamble staff on real product-development tasks. People working alone with AI produced ideas rated as highly as those of two-person teams without AI, and did it faster.

By
Last reviewed
Corrections
Report an error
Outcome Measured result

Researchers measured the result and published how.

In the guide Enterprises

At a glance

Company
Procter & GambleConsumer goods manufacturer
Size
Enterprise (1,000+ employees)
Industry
Manufacturing The work studied was product development, not production on the factory floor.
Where and when
Europe and the Americas. May–July 2024.
Who produced the figures
Measured by researchersThe figures come from a working paper by researchers at Harvard, Wharton and other business schools. Four of the eleven authors work at Procter & Gamble. The paper has not been peer-reviewed.

What they did

Procter & Gamble ran one-day online workshops for 776 of its staff. Each person worked on a real product problem from their own business unit: baby care, feminine care, grooming or oral care. Participants came from two groups: research and development, and commercial roles.

People were assigned by chance to one of four set-ups: alone, alone with AI, in a pair, or in a pair with AI. A pair had one person from each group. The AI was a GPT-4 tool. Those who used it had one hour of training first. Expert judges then scored the quality of each proposal. They did not know which set-up had produced it.

The numbers

Quality is measured against people working alone without AI. The unit is the standard deviation, a common measure of how far a result sits from the average.

Set-upQuality compared with one person, no AITime spent
Pair, no AI0.24 higherAbout the same
One person with AI0.37 higher16.4% less
Pair with AI0.39 higher12.7% less
  • Best ideas. Pairs with AI were 9.2 percentage points more likely to produce a proposal rated in the top 10%. Among people working alone without AI, 5.8% reached that level.
  • Less divided by department. Without AI, research staff proposed technical ideas and commercial staff proposed commercial ones. With AI, both produced balanced proposals.
  • Mood. People who used AI reported more positive feelings about the work.

What the case shows

  • AI helped on idea-stage product work. The gain was measured in an experiment with a comparison group, inside a working manufacturer. Judges rated proposals. No product was launched and sold as part of the study.
  • One person with AI did about as well as a pair without it. The paper describes the two sets of scores as very similar.
  • Little training was needed. One hour, for staff the authors describe as fairly new to prompting.

Other experiments show that AI helps on some tasks and harms on others. See the study of 758 consultants.

What this says about the practices

Practices the company applied

Sources

  1. Dell’Acqua, Ayoubi, Lifshitz, Sadun, Mollick, Mollick, Han, Goldman, Nair, Taub and Lakhani (2025). The Cybernetic Teammate: A Field Experiment on Generative AI Reshaping Teamwork and Expertise. NBER Working Paper 33641. nber.org
  2. Dell’Acqua, McFowland, Mollick, Lifshitz-Assaf, Kellogg, Rajendran, Krayer, Candelon and Lakhani (2026). Navigating the Jagged Technological Frontier: Field Experimental Evidence of the Effects of Artificial Intelligence on Knowledge Worker Productivity and Quality. Organization Science, 37(2), 403–423. What this study did and found
  3. Vaccaro, Almaatouq and Malone (2024). When combinations of humans and AI are useful: A systematic review and meta-analysis. Nature Human Behaviour, 8(12), 2293–2303. What this study did and found