Page 497 - ISC PROCEEDINGS 21.4

standardizing years of historical data. Furthermore, banks must navigate the delicate
balance between leveraging data for AI training and complying with stringent personal
data protection regulations.
Shortage of High-Quality AI Talent: The Vietnamese market is experiencing a severe
deficit of professionals capable of bridging the gap between financial expertise and AI
technology. Commercial banks are not only competing with one another but are also
locked in a 'war for talent' against Big Tech giants (such as Google, Grab, and Shopee) to
attract top-tier Data Scientists. Simultaneously, traditional banking staff face immense
pressure to undergo reskilling to operate alongside automated systems, leading to a
heightened risk of personnel turnover.
Black Box' Risks and Transparency: The lack of interpretability in AI models poses a
significant risk; if a loan is rejected, banks struggle to provide a detailed 'why' to
customers or regulators, which undermines transparency in the financial sector.
Furthermore, if training data contains inherent algorithmic bias—such as historical
defaults concentrated in specific regions—AI may inadvertently 'learn' and perpetuate
discriminatory practices against those demographics in the future.
Legacy Infrastructure Barriers: A majority of Vietnamese banks still operate on
legacy Core Banking platforms, making it exceptionally difficult to integrate modern AI
applications that require real-time processing into bulky, monolithic systems. The
financial burden of migrating infrastructure from on-premise servers to the cloud to
support AI is a substantial hurdle, particularly for mid-sized institutions.
Regulatory Vacuum and Cybercrime: There is currently no comprehensive legal
framework to assign liability when AI causes errors or financial losses for customers.
While banks deploy AI for security, cybercriminals are simultaneously utilizing AI to create
Deepfakes (facial and voice spoofing) to bypass eKYC layers, exerting constant defensive
pressure on financial institutions.
Table 4. Typical challenges faced by individual banks
Bank Key Challenges
Pressure to maintain system stability amidst surges in
TPBank and MB
automated transaction volumes.
Enormous investment costs for Cloud infrastructure
Techcombank
and high-end customer data security.
Controlling risks within automated loan approval
VIB and VPBank
models during adverse market fluctuations.
Source: Compiled by the author from bank reports
3.3. Strategic solutions for Vietnamese banks in the AI era
To overcome barriers related to data, human resources, and algorithmic risks,
leading banks such as TPBank, MB, Techcombank, VIB, and VPBank are implementing a
comprehensive set of strategic solutions with high perforAlgorithmic Risksmance
expectations (see Table 5).
Table 5. Matrix of Challenges, core solutions, and expected outcomes
Challenges Core Solutions Expected Outcomes
Synchronized data with
Data Junk / Implementing Data Lakes and real-time analytics
Fragmentation migrating to Cloud Computing.
capabilities.
Upskilling current staff and A workforce proficient
Talent Shortage
recruiting for Agile job roles. in both Finance and AI.

496