A “dual-purpose” teammate: working with GenAI

Frame of the research. Human-AI teams hold substantial promise due to the complementary strengths of human cognition and artificial intelligence. Despite this potential, a growing body of research suggests that AI-augmented teams do not consistently deliver the expected performance and in some cases contribute to undermine core team properties introducing negative aspects that may adversely affect employee well-being, including psychological distress and depression (Kim et al., 2025). A critical but underexplored question concerns whether these outcomes depend not merely on AI access, but on the structural properties of the human team through which AI is used. Purpose of the paper. This study asks: (RQ1) how do team size and seniority composition influence performance outcomes in AI-enabled task execution? And (RQ2) how do team size and seniority composition shape emotional dynamics in the same context? We argue that the introduction of GenAI as a broadly capable collaborator alters the mechanisms through which team diversity traditionally confers its advantage — shifting the primary performance bottleneck from idea generation to coordination and integrative execution. Methodology. To address these questions we conducted an experiment within Accenture, where consultants were randomly assigned to teams consisted of either a single individual or two individuals working with AI, who could hold the same or different roles within the organization. Participants received a use case to be developed throughout the day. Each solution submitted received three independent single-blind evaluations from different experts to reduce evaluation bias and increase reliability. Evaluators rated each solution on five dimensions: quality, impact, novelty feasibility, and technical sophistication. Results. Our findings indicate that homogeneous teams, characterized by the same level of seniority, outperformed both individuals and heterogeneous teams, attaining the highest levels of feasibility. In contrast, heterogeneous teams did not exhibit statistically significant performance advantages relative to individuals. Beyond performance outcomes, teamwork was associated with more favorable emotional experiences. Participants working in teams reported lower increases in frustration and stress compared to individuals, for whom both measures rose substantially. These results suggest that collaborative settings can mitigate negative emotional responses connected with working with Generative AI. Managerial implications. These findings reframe AI as a dual-purpose organizational tool: one that simultaneously enhances team performance and buffers workers against the negative emotional consequences of AI-enabled work — but whose benefits are strongly contingent on seniority composition. Organizations should treat team structure, not merely AI access, as a strategic lever for realizing AI's competitive value. Research limitations. We also acknowledge two key limitations. First, our experiment relied on one-day virtual collaborations, which do not fully capture the day-to-day complexities of organizational. Second, we examined small, two-person teams specifically constructed for the experiment. These factors may influence both performance and emotional dynamics in ways that our experimental design could not fully replicate. Originality of the paper. This study makes three contributions. First, it qualifies the diversity–performance relationship under AI augmentation, showing that seniority homogeneity outperforms heterogeneity when GenAI externalizes ideational variety. Second, it extends exploration–exploitation theory to hybrid human-AI configurations, arguing that AI decouples the two functions by handling exploration and leaving convergence to the human team. Third, it advances human–AI teaming theory by demonstrating that seniority composition is a first-order boundary condition for AI-enabled team performance, a dimension largely absent from prior experimental designs.