Compared with other types of artificial intelligence (AI), the most distinctive feature of generative AI is its “creative” nature. Knowledge is a crucial factor in this interactive process; thus, the knowledge creation process is inevitably and profoundly affected. Existing theories are insufficient to explain how generative AI reshapes knowledge creation. This study focuses on AI chip startups – arenas characterized by breakneck innovation but fragile trust – and interrogates the black box that connects AI affordances (transparency, traceability and dynamic feedback) to trust trajectories and, in turn, to knowledge exposure, validation and orchestration. Anchored in knowledge-based theory, this study aims to develop a transferable process model that can explain the mechanisms involved in this context, thus redefining the boundaries of knowledge management in the AI era.
This study features a multicase study methodology, tracking multiple chip startups within high-tech clusters that embed generative AI into their R&D, design and customer solution workflows. Fifty semistructured interviews (lasting a total of 3,115 min) produced 376 transcripts with a volume exceeding 500,000 words and was complemented by participant observations in shared labs and archival records (including code repositories, chat logs and version histories). The data analysis process followed rigorous multicase research procedures and systematically incorporated the methods associated with grounded theory. The analytical process involved two core steps – within-case analysis and cross-case analysis – guided by theoretical sampling and iterative constant comparison, thereby ultimately generating findings with high levels of validity and reliability.
The core findings of this study can be summarized in three progressive layers. First, this study presents the construction of a process model of generative AI integration into startups. Second, building upon this process model, a novel knowledge creation model is developed that features three phases: knowledge exposure, knowledge validation and knowledge orchestration. This model systematically explains how generative AI disrupts and reshapes the traditional activities of socialization, externalization, combination and internalization, thereby transforming knowledge creation into a dynamic process of deep human–AI collaboration. Third, from the perspective of trust, this study provides an in-depth analysis of the mechanisms underlying this new knowledge creation process. It demonstrates how trust evolves through three key mechanisms – building trust through transparency and traceability, enhancing trust through contextual alignment and collaborative optimization, and stabilizing trust through dynamic feedback. These findings reveal that trust under conditions of technological mediation constitutes the central driving force catalyzing a paradigm shift in knowledge creation.
The originality and value of this study are reflected at two levels. First, this study transcends the instrumental perspective, which views AI merely as a tool, by constructing a new knowledge creation process model comprising knowledge exposure, knowledge validation and knowledge orchestration, thereby systematically explaining how generative AI fundamentally reshapes knowledge activities. Second and more innovatively, this study uses trust as its central analytical lens for an in-depth examination of the mechanisms underlying the new knowledge creation process. This study reveals how the technological attributes of AI, such as transparency, traceability and dynamic feedback, drive the evolution of knowledge creation by building, enhancing and stabilizing trust. Accordingly, this study fills a theoretical gap in knowledge management regarding trust mechanisms in the AI era and offers new conceptual tools and actionable insights that can support enterprise practice.
