AI paradox in Nepalese banking: operational efficiency vs. ethical and regulatory risks

Abstract

The rapid integration of artificial intelligence (AI) in Nepal's banking sector presents a critical paradox. While adopted to enhance operational efficiency, its implementation often outpaces the development of essential ethical and regulatory frameworks. This empirical study of 400 banking professionals across 28 commercial banks identifies a four-construct framework—AI infrastructure, model governance, service integration, and ethics capacity—that influences performance. Regression analysis reveals that ethics training (β = 0.216, p < 0.001) and AI-enabled services (β = 2.012, p < 0.001) significantly boost operational performance. Conversely, opaque model governance (β = -0.860) and subpar infrastructure (β = -0.788) detrimentally affect efficiency. The findings suggest that hybrid governance systems and regulatory sandboxes can bridge this implementation gap by striking a balance between innovation and responsibility. This study contributes to the understanding of AI adoption in resource-constrained environments, offering critical insights for financial institutions and policymakers.