As artificial intelligence (AI) models become increasingly widespread, a bank's competitive edge is no longer determined by technology alone, but by the quality of its data. The more complete and accurate the data, and the more effectively it is managed, the greater the value AI can generate for business operations.
In just a few years, AI has evolved from an experimental technology into an “assistant” for many banks, supporting everything from customer service, credit approval, and fraud detection to risk management. However, as the technological gap between institutions narrows, a new question is emerging: What will determine banks' competitive advantage in the AI era?
The answer does not lie in which bank possesses the most advanced AI model. In reality, most AI models today are built on shared technological platforms that can be adopted by multiple organisations. What sets banks apart is their data – a unique asset accumulated over decades of operations, ranging from transaction histories and customer behaviour to business processes and risk-management experience.
According to Dao Minh Tu, Standing Vice Chairman and Secretary-General of the Vietnam Banks’ Association, AI can only deliver its full potential when powered by high-quality data and supported by a robust data governance framework. This is also essential for banks to use the technology safely and transparently and generate tangible value, rather than merely showcasing technological applications.
Sharing this view, Nguyen Phuong Thanh, deputy director of banking and financial services advisory at Ernst & Young Vietnam (EY Vietnam), said the banking industry is entering a new phase, moving from applying AI to individual business functions towards building banks that operate on an AI foundation.
Differences in data will ultimately determine each institution's ability to personalise services, manage risks and harness AI effectively.
According to EY Vietnam, the transition to AI-powered banking requires institutions to establish a comprehensive data governance framework covering data quality, master data, metadata and data lineage. This is not merely a technological requirement, but also the foundation for AI to operate transparently, reliably and at scale across multiple professional areas.
From a practical implementation perspective, Bui Gia Hieu, Head of Security at Military Commercial Joint Stock Bank (MB), said that rather than giving AI full decision-making authority, the bank had built an end-to-end operating model connecting cybersecurity threat intelligence sources with layers of defence throughout the digital transaction journey.
Under this model, AI processes large volumes of data, including data collection, filtering, enrichment, event correlation and investigation recommendations, while final decisions remain in the hands of experts. Practical implementation has demonstrated clear benefits, with AI helping MB increase its capacity to receive and analyse cybersecurity threat intelligence from around 3,000 to 10,000 items per day, more than three times the previous level.
Notably, MB does not measure the effectiveness of AI by the number of models or deployment licences, but by its ability to shorten the detection-testing-alert-response cycle. AI is used to process data at a higher speed and larger scale, while any action affecting systems or customers must still go through human approval mechanisms.
At the same time, as AI continues to develop, data governance requirements are becoming increasingly closely linked to the legal framework.
According to Hoang Ha, a data protection and AI governance expert and CEO of Data Protectify, data use and AI applications must comply with the laws on data and on personal data protection as well as requirements concerning risk management, transparency and accountability. These are not merely compliance requirements, but also important factors in strengthening customers' trust in digital services.
