Can AI effectively replace a human teammate?
Researchers from Harvard, Wharton, Procter & Gamble and other institutions did an experiment to find out. They asked 776 professionals from P&G to come up with new product ideas and measured their performance. Half used artificial intelligence (AI) and the other half didn’t.
The results? “AI significantly enhances performance: Individuals with AI matched the performance of teams without AI, demonstrating that AI can effectively replicate certain benefits of human collaboration,” the researchers said in a March 2025 research paper. “Moreover, AI breaks down functional silos.”
When it comes to effectiveness, employees and teams using AI were also three times more likely to suggest product ideas in the top 10% of solutions, based on a ranking by human experts. This can translate to real revenues.
The paper’s findings carry implications for optimal team sizes and team make-up at organizations. It could reshape traditional teamwork and redefine how expertise is shared in the workplace.
The study, “The Cybernetic Teammate,” conducted workshops among P&G professionals from May to July 2024. The authors measured performance gains, broadening of expertise and level of social engagement.
What made this study especially meaningful was its real-world setting. The experiment embedded AI directly into the company’s actual product development process — it was not made in isolated tests. Participants were tasked with solving real business challenges across actual brand categories.
P&G’s global brands include Pampers, Head & Shoulders, Old Spice, Olay, Gillette, Oral-B, Braun, Olay and many more.
The study’s findings could be unsettling for workers. Among those who use generative AI weekly, 74% said it can automate or replace elements of their job, according to a January 2025 PYMNTS Intelligence report.
However, not all research agrees with the Harvard study. According to an October 2024 MIT paper, human-AI collaborations do not always lead to outperformance compared to humans or AI acting alone. Reasons for this counterintuitive finding include communication barriers, trust issues, ethical concerns and the need for effective coordination between humans and AI.
Read more: AI’s Next Job Is as Your Collaborative ‘Digital Chief of Staff’
Inside the ExperimentThe researchers created four types of teams: individuals working alone without AI, teams of two working without AI, individuals working with AI, and teams of two working with AI.
P&G employees were asked to come up with new product ideas in four lines of business: Baby care, feminine care, grooming and oral care. The target markets were the Americas and Europe.
Their solutions were graded by human experts for quality. The generative AI tool used was built on OpenAI’s GPT-4 foundation model accessed through Microsoft Azure cloud. The experiments took place from May to July 2024.
The researchers discovered that one person using AI could replicate the work of two-person teams not using AI. That means one person paired with gen AI produced ideas and solutions of equal quality to those generated by a traditional human duo.
But the implications go beyond productivity.
At P&G, cross-functional collaboration between R&D and Commercial divisions is a foundational part of product innovation. However, each group tends to gravitate toward solutions in their expertise: R&D favoring technical solutions, Commercial leaning toward market-friendly ideas.
When employees used AI, those silos dissolved. Commercial experts suggested more technical ideas, and R&D specialists incorporated more marketplace ideas. AI served to expand access to specialized knowledge.
This finding can be a revelation to companies dealing with knowledge silos.
“AI serves not just as an information provider but as an effective boundary-spanning mechanism, helping professionals reason across traditional domain boundaries and approach problems more holistically,” the authors said.
Breaking down the results:
“We found that individuals randomly assigned to use AI did as well as a team of two without AI,” co-author and Wharton professor Ethan Mollick concluded, in a LinkedIn post. “AI-augmented teams produced more exceptional solutions. The teams using AI were happier as well.”
Thus, the standard view of AI as productivity tool “may be too limiting. Today’s AI can function as a kind of teammate, offering better performance, expertise sharing, and even positive emotional experiences,” Mollick said.
For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.
The post How Generative AI Affects Teamwork at Companies appeared first on PYMNTS.com.