INTERNATIONAL RESEARCH / PEER-REVIEWED ARTICLE

Can AI replace a teammate?

A field experiment with 791 P&G professionals found that AI-assisted individuals approached two-person teams on some product-innovation tasks. We explain where human selection retained value.

ACADEMIC WORK REVIEWED

The Cybernetic Teammate: A Field Experiment on Generative AI and Teamwork

ResearchersFabrizio Dell’Acqua, Charles Ayoubi, Hila Lifshitz, Raffaella Sadun, Ethan Mollick, Lilach Mollick, Yi Han, Jeff Goldman, Hari Nair, Stew Taub and Karim R. Lakhani.

Organization Science · 12 June 2026 · 37(4):1217–1242 · Peer-reviewed field experiment · Source language: English

Open the original publication Find on Google Scholar

ENGLISH REVIEW AND COMMENTARY: DIJITALPI

We explain the paper in plain English, discuss what it may mean for organisations and identify our own examples separately.

Read the explanation Go to 3 business examples

English review: 10 September 2026 · The source date appears in the citation above.

Published review

Source and editorial checks are complete. DijitalPi did not conduct a customer experiment for this paper.

LET'S READ THE RESEARCH TOGETHER

We first explain the researchers' question, method and findings. We then discuss how to interpret the results, clearly separating DijitalPi's commentary from the source.

01 / WHAT DID THE RESEARCHERS WANT TO UNDERSTAND?

Which functions of human teamwork can AI provide in product innovation?

Dell’Acqua and colleagues tested AI’s role in performance, expertise integration and work experience inside a real organisation.

Explaining the context · DijitalPi commentary

Teamwork combines expertise, selection and accountability, not only the labour of two people.

The study treats conversational AI as a potential collaborator and tests the limits of that analogy through real task outcomes.

02 / HOW WAS THE RESEARCH CONDUCTED?

How did they test the question?

At P&G, 791 professionals working on real product-innovation problems were randomized to work alone, in a two-person team, alone with AI or in a team with AI during a one-day remote workshop.

Understanding the method · DijitalPi commentary

Commercial and R&D staff allowed the researchers to measure cross-functional knowledge integration. Independent evaluators reviewed the output.

A short workshop strengthens causal comparison but does not measure long-term team relationships, implementation or market success.

03 / RESEARCH FINDINGS

AI-assisted individuals approached two-person teams on some quality measures.

Solutions from AI-assisted individuals reached quality similar to teams without AI. AI narrowed the gap between commercial and technical expertise. AI-assisted teams were also strong among top-scoring outcomes, while human judgement retained value in idea selection.

Professionals: 791P&G commercial and R&D employees.
Conditions: 4Individual or team, with or without AI.

Source: Organization Science article and DOI ↗

04 / DIJITALPI'S EXPLANATION

How should we interpret these findings?

AI can help one person consider questions from another function without assuming that function’s accountability.

Generating ideas and selecting the right one are separate skills. Human evaluation may remain critical for portfolio decisions.

Team-size decisions need implementation, accountability and long-term learning measures in addition to average quality.

05 / CONCLUSION AND OPEN QUESTIONS

What did we learn, and what do we still not know?

AI provided some performance and knowledge-integration benefits of teamwork in specific innovation tasks. A one-day remote workshop does not prove that permanent teams are generally unnecessary.

What happens to quality, speed and market outcomes when AI-assisted ideas enter a real product pipeline? The study did not measure that long-term result.

Return to the researchers' original publication

Organization Science article and DOI

DIJITALPI'S APPLICATION COMMENTARY

What could this look like in your organisation?

We created these scenarios to make the topic concrete. They are not cases from the paper or measured client results.

EXAMPLE 01

One person enters two fields

A marketer uses AI to write a product idea with a strong technical component.

Does that complete technical approval?

Open the recommendation for example 01

Separate breadth of ideas from expert verification.

AI can combine the language of different functions. Leave safety, cost and feasibility decisions to the responsible experts.

EXAMPLE 02

Many ideas, weak selection

The AI-assisted group produces stronger ideas but struggles to choose which one to implement.

Does better generation guarantee better selection?

Open the recommendation for example 02

Write the selection criteria in advance.

Score customer need, technical risk, cost and delivery time separately. Treat AI output as a candidate pool rather than the final decision.

EXAMPLE 03

A one-day workshop becomes a permanent design

Management sees the experiment and plans to shrink a long-term product team.

Does a short study prove a permanent team design?

Open the recommendation for example 03

Preserve the study’s time and relationship limits.

The work took place remotely in one day on specific product problems. Measure long-term learning, accountability and implementation before making a permanent structural decision.

TRY IT WITH YOUR TEAM

Redraw team work in three stages.

Separate idea generation, selection and implementation. For every stage, name the AI role, accountable person and criterion for moving forward.

Explore our strategic marketing consulting

Source and ownership

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.
Harvard Business School; HBS AI Institute; ESSEC Business School; Warwick Business School; The Wharton School; Procter & Gamble.

Organization Science · 12 June 2026 · 37(4):1217–1242 · Peer-reviewed field experiment. Open the original publication · Find on Google Scholar

The academic work belongs to the researchers named above. This page contains DijitalPi's explanatory review and original business examples; it is not a full translation of the paper.

← Return to all research articles