Answers are grounded in this paper and cite their sources.
Generative AI systems are increasingly integrated into collaborative team settings, yet the field lacks a structured account of the roles these systems occupy relative to human members. This systematic literature review examines how generative AI has been conceptualized and studied as a team participant across organizational, educational, and technical domains. Following PRISMA 2020 guidelines, a search of the Scopus database retrieved 7,541 records published between 2018 and 2026. After title and abstract screening, 370 records were advanced to full-text assessment; 65 studies met all eligibility criteria and provided codeable role assignments. Seven role categories emerged from inductive content analysis: Partner (n = 44, 67.7%), Assistant (n = 8, 12.3%), Tool (n = 7, 10.8%), Autonomous Member (n = 3, 4.6%), Evaluator (n = 1, 1.5%), Decision-maker (n = 1, 1.5%), and Other (n = 1, 1.5%). The Partner category dominates the corpus, reflecting a broader conceptual shift in which AI is framed not merely as an instrument but as a contributor with social standing in the team. Experimental and quasi-experimental designs account for 31.3% of codeable studies, indicating a maturing evidence base, although qualitative and conceptual work remains substantial at 40.3% combined. A notable gap is the low granularity with which studies specify the process stage at which AI participation occurs: only 16 of 65 studies identified a specific substage beyond the generic label of workflow. Theoretical and practical implications for human-AI team design, trust calibration, and role governance are discussed.
No takes yet. Share an insight, caveat, or question.
Melo et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: