Idea Generation With a Majority of AI Teammates: Exploring Productivity Effects, Process Gains and Process Losses

Leveraging AI agents as teammates has gained momentum in research and practice. Yet, how human team members react to multiple AI teammates (AITMs), particularly when they are the majority, is underexplored. Therefore, we conducted an exploratory study applying experimental methods, in which 66 participants collaborated in teams of three humans and four AITMs on an idea generation task. Our results show that the AITMs were highly productive, with half of the top ideas being AI-generated, and contributed to process gains for human team members. However, we also observed process losses, such as attention blocking or reduced human self-efficacy. While participants were open to future collaboration with multiple AITMs, they preferred team compositions with an equal or fewer number of AITMs than humans. We conclude that an optimal team composition balances productivity gains with the need to mitigate process losses that arise in AI-majority teams and maintain human engagement.