A study involving 776 Procter & Gamble experts points to a practical shift in how companies may think about AI at work. In a controlled product development setting, people using AI were able to close the gap with traditional two-person teams.
The finding is not a simple claim that AI replaces collaboration. The results show gains in speed, detail and idea quality, but they also come with limits that matter for real enterprise work.
What the workshop tested
The research was built around one-day workshops in which participants created product ideas for various P&G business units. Some people worked alone, while others worked in teams made up of one commercial expert and one technical expert.
Half of both the teams and the individuals received access to GPT-4 and GPT-4o. That design allowed the researchers to compare individual work, team work, AI-assisted individual work and AI-assisted team work inside the same general task environment.
The baseline result was clear: teams without AI outperformed individuals by 24%. That fits the normal expectation that combining different kinds of expertise should produce stronger work than relying on one person.
But AI changed the comparison. Individuals using AI improved by 37%, which brought them up to the performance level of non-AI teams. In other words, a solo expert with AI support was able to match what a traditional two-person team achieved without AI in this setting.
Teams still gained, but the picture is more nuanced
The best overall performance came from teams that also had AI support. Those groups reached a 39% improvement, higher than the other formats tested.
Still, the difference between AI-supported teams and AI-supported individuals was not statistically significant. That detail matters because it suggests the main performance gap may not simply be about headcount. Access to AI changed the shape of the work enough that one person could perform near the level of a larger unit.
There was one area where teams with AI stood out more sharply. They were about three times more likely to produce solutions that ranked in the top 10% of quality scores. So while individuals with AI could match average team performance, the strongest outcomes were more likely when human collaboration and AI support were combined.
The study also found that groups using AI worked 12-16% faster while producing longer and more detailed solutions. More than 75% of AI-generated content was kept by a "substantial proportion" of the groups, according to the study.
AI appeared to broaden expertise
One of the most interesting findings was not just that AI improved output. It also seemed to reduce differences between types of experts.
Without AI, technical experts tended to stay focused on technical solutions, while sales experts concentrated on market aspects. With AI assistance, both groups produced more rounded proposals.
That points to a possible role for AI as a bridge between professional perspectives. It does not mean every suggestion was equally valuable, or that expertise no longer mattered. But within the workshop, AI appeared to help participants move beyond the limits of their usual specialty.
The effect was especially visible among less experienced product development employees. Without AI, they performed relatively poorly, including when they worked in teams. With AI support, their results rose to levels comparable with teams that included experienced members.
The emotional response was also notable. Participants using AI reported more positive emotions such as enthusiasm and energy, with fewer signs of anxiety and frustration. That runs against the assumption that introducing new technology necessarily adds stress for users.
Why the “AI teammate” idea needs caution
The researchers argue that companies should think about AI as more than a productivity tool. They suggest it can be viewed as an additional team member because it appears to support performance, expertise sharing and user experience.
Study leader Ethan Mollick framed the point this way: "Although it is not human, it replicates the core benefits of teamwork - improved performance, sharing of expertise and positive emotional experiences. This team player perspective should lead organizations to think differently about AI."
That interpretation is important, but it also has limits. The study mainly used AI as a chatbot, which may have restricted what the technology could do. Chat formats can be useful for fast ideation, but the quality of the result still depends partly on chance and heavily on the user's expertise.
There is also an evaluation problem. AI can generate a large amount of text, and that output needs professional judgment. A marketer may not be able to assess technical suggestions properly, for example, and could end up trusting AI too easily. Producing more words is not the same as producing useful knowledge.
The research also took place in a one-day workshop. That format does not reflect the full complexity of enterprise work, where ideas usually move through repeated cycles over longer periods. It remains unclear how much AI-generated material would survive those extended development processes.
The source article also raises another possible direction: in product development, where structured processes already exist, AI may be more useful when it helps standardize those processes rather than when it is humanized as a conversation partner. Companies could then make AI-enhanced processes, or their outputs, available to teams.
What companies should take from it
The study gives companies a reason to revisit how they assign AI inside teams. It suggests that AI can help individuals perform at a higher level, help teams move faster and make proposals more detailed.
But implementation will not be uniform. The source notes that success varies based on corporate culture, existing workflows and technical infrastructure. Those factors shape whether AI becomes a useful support system or just another layer in the process.
The biggest open question is what happens over time. AI integration may affect skill development and knowledge transfer, both of which are crucial for maintaining competitive advantage. A short workshop can show what AI changes in the moment, but it cannot fully answer what it does to organizational learning.
For now, the practical takeaway is balanced. AI can make a single expert look more like a small team in certain product development tasks. But the strongest and most durable value may depend on how well companies pair AI with human review, structured workflows and real expertise.