Governing generative AI in R&D: process agility as a dynamic capability for exploratory innovation

Purpose This study investigates the structural mechanism through which firms translate AI-enabled synergistic invention capabilities (AI-SIC) into exploratory innovation, addressing the governance challenges of algorithmic variance within digitalized R&D architectures. Design/methodology/approach Grounded in the dynamic capabilities view (DCV), the study employs a quantitative cross-sectional design. Data were collected from 328 high-tech R&D executives (CTOs, CIOs and R&D Directors) in Taiwan and analyzed using Partial Least Squares Structural Equation Modeling (PLS-SEM). Findings The empirical results reveal a full mediation mechanism. Raw AI-SIC does not directly yield exploratory innovation; rather, its strategic value is entirely channeled through R&D process agility. This agility acts as a crucial cognitive filter, enabling organizations to mitigate stochastic algorithmic variance and potential technological hallucinations. Originality/value This research extends the DCV into the algorithmic era by shifting the scholarly focus from mere digital technology adoption to internal variance governance. It conceptualizes the novel multidimensional AI-SIC construct and reconceptualizes R&D process agility as an indispensable dynamic capability for navigating generative AI. Crucially, this study contributes primarily to Dynamic Capabilities research while informing the digital innovation literature.