When Does Generative AI Adoption Pay Off? Ethical Governance as a Dynamic Capability in SMEs

Generative AI (GenAI) offers substantial innovation opportunities for small and medium-sized enterprises (SMEs), yet its performance benefits vary considerably across firms. Drawing on the resource-based view (RBV) and dynamic capabilities perspective, this study theorizes ethical AI governance as a capability bundle—comprising transparency/explainability (TE), fairness/bias control (FB), privacy/data protection (PD), and accountability/responsibility (ACCT)—that conditions the innovation payoff from GenAI adoption. Using 218 usable manager responses from Canadian SMEs, we test the proposed model using covariance-based structural equation modeling in AMOS 22 and examine moderation through multigroup analysis with measurement-invariance checks. The results show that GenAI adoption is positively associated with innovation performance, while TE, PD, and ACCT significantly strengthen this relationship. In contrast, FB exhibits a significant negative moderating effect, a pattern consistent with a short-run maturity-stage trade-off in which overly rigid or poorly integrated fairness controls may slow early experimentation in resource-constrained SMEs, even though fairness assurance remains essential for responsible scaling. The findings also reveal asymmetric adoption drivers: green-innovation orientation is the strongest positive driver, whereas carbon-footprint pressure and economic disruption are negatively associated with adoption. Higher innovation performance, in turn, is positively associated with adaptability and resilience, operational efficiency, and revenue growth. Overall, the study reframes ethical AI governance from compliance overhead to a dynamic capability and identifies governance-related boundary conditions that help explain when GenAI adoption is associated with stronger innovation performance and downstream firm outcomes in SMEs.