Toward a generative architecture for human–AI collaboration: A systematic review-based microfoundational framework for value creation

PURPOSE: This study aims to decipher the black box of Human–AI collaboration in the process of organizational value creation. To this end, moving beyond technology acceptance models, it develops the Determinant-Mechanism-Outcome (D–M–O) model, an integrated process architecture, presented not as an empirically tested model but as a conceptual framework synthesized from a systematic literature review. The research focuses on explaining how organizational antecedents are transformed into strategic performance through specific micro-processes by conceptualizing Human–AI collaboration as a “generative architecture” driven by the synergy of human wisdom and machine computation. METHODOLOGY: Adopting a conceptual design based on a systematic literature review, the study analyzes 92 empirical articles through the PRISMA protocol. To synthesize fragmented findings, 261 unique study-level dyadic relationships (D–M, M–M, M–O), obtained after removing within-study duplicate dyads, were identified. Network-based centrality analysis was applied using the data derived from thematic and relational frequency analyses and these dyadic relationships. FINDINGS: The findings reveal that 91.95% of the examined relationships involve mediating mechanisms, challenging the traditional direct effect assumption in the literature. While Trust in AI emerged as the pivot node of the system with the highest centrality score, Collaboration Intensity with AI and Hybrid Intelligence mechanisms repeatedly emerged as core mechanisms across the reviewed studies. The study identifies three recurring value creation routes focused on the Design-Mediated Trust (Path I), Strategic Capability (Path II), and Learning-Supported Competency (Path III), suggesting that value may be constructed through the coordinated interaction of these mechanisms within a cumulative architecture. IMPLICATIONS: Theoretically, the study aims to offer a holistic framework for Human–AI collaboration by synthesizing Socio-Technical Systems, Complementarity Theory, and Micro-Foundations approaches. At a practical level, it suggests that managers may help mitigate stressors such as algorithmic pressure and job insecurity primarily through trust-building practices and by strengthening employees’ sense of control over AI systems. ORIGINALITY & VALUE: This research contributes to the literature by reframing the guiding question from “what is AI used for?” to “through which micro-mechanisms does AI generate strategic value?”. Furthermore, by introducing the generative architecture concept, it offers an alternative to technological determinism, suggesting that Human–AI collaboration may be better understood as a dynamic, mechanism-driven process rather than a static input–output relationship. In this respect, the study offers a conceptual starting point for future theorizing on intelligent organization design.