ARTIFICIAL INTELLIGENCE ADOPTION IN VIETNAM’S BANKING SECTOR: CURRENT PRACTICES, CONSTRAINTS, AND SOLUTIONS

Abstract

Artificial intelligence (AI) is moving from isolated experimentation toward more embedded use in Vietnam’s banking sector, but the depth of adoption and the quality of public performance evidence remain uneven. This article examines current practices, implementation constraints, and governance priorities through a structured qualitative review updated to 24 August 2026. The review combines peer-reviewed banking research, international risk-governance guidance, Vietnamese legal and supervisory documents, and recent official disclosures from major commercial banks. The evidence indicates that digital identity and biometric authentication are operational and strongly regulation-driven, while machine-learning applications in underwriting, personalization, fraud monitoring, and decision support are increasingly embedded at leading banks. Generative AI has also moved beyond experimentation in several institutions, although sector-wide maturity remains uneven. Recent disclosures provide more concrete operational metrics than earlier sources, but most remain self-reported and cannot establish causal effects on profitability or risk. The main constraints concern fragmented data and legacy integration, model and conduct risk, cybersecurity and third-party dependence, uneven governance capacity, and shortages of interdisciplinary expertise. The article applies a technology–organization–environment framework with a cross-cutting AI-risk governance layer and proposes a sequenced implementation matrix for banks and supervisors. The central implication is that the value of AI in banking depends less on acquiring models than on integrating data, controls, accountable ownership, and human oversight throughout the AI lifecycle.