Page 69 - ISC PROCEEDINGS 21.4

Third, real-time operation. Disbursement data and project progress should be
continuously updated, enabling real-time online dashboards instead of relying solely on
quarterly or annual reporting. This aligns with the proactive fiscal governance approach
recommended by the World Bank in its 2022 Public financial governance reports.
Fourth, robust security and strict access control. In the context of rising global
cyberattacks, the system must comply with high-level information security standards,
implement function-based access control, and maintain comprehensive audit trails of all
data operations.
In the long term, the Smart Treasury Platform should serve not only as an
expenditure control tool but also as an analytical foundation for evidence-based fiscal
policymaking.
3.5.2. Integrating AI into the public finance ecosystem
Once an integrated data platform has been established, the deployment of AI
should follow a cautious and well-controlled roadmap.
First, developing a public investment project risk-scoring model. Based on historical
data on project progress, adjustments to total investment, contractor changes, and audit
results, ML models can estimate the probability of delays or cost overruns. According to
estimates by the Organisation for Economic Co-operation and Development (OECD),
applying data-driven risk management can reduce losses caused by poor governance in
the public sector by approximately 10-15%.
Second, network analysis in public procurement. AI can detect abnormal
relationships among contractors, high concentration in contract allocation, or repetitive
bidding patterns that signal potential collusion. This is particularly important in the
context where PIE accounts for a significant share of aggregate demand in the economy.
Third, applying natural language processing (NLP). The system can automatically
analyze thousands of pages of contracts, adjustment appendices, and progress reports to
extract risk-related information. According to a survey by McKinsey & Company,
automating document analysis can save 20-30% of administrative processing time.
Fourth, ensuring the principle of “AI as support - humans as decision-makers.”
Recommendations generated by the system should serve only as references, while
decisions regarding inspections or ST adjustments must be made by authorized officials.
At the same time, Viet Nam should develop regulations on algorithmic transparency, AI
system audits, and personal data protection in line with international best practices.
3.5.3. Developing human resources for data analytics
Technology cannot deliver its full potential without a workforce capable of
operating and leveraging data effectively. According to the Vietnam Digital Economy
Report by Google and Temasek, demand for digital talent in Viet Nam is growing at an
average annual rate of 20-30%, while the supply of highly qualified professionals remains
limited.
First, reskilling financial and treasury officials. It is essential to retrain finance and
treasury staff in statistics, BDA, and quantitative thinking. Training programs should
combine theoretical instruction with hands-on practice using real sectoral data.
Second, establishing dedicated BDA Units within the public financial system. These
units would be responsible for developing analytical models, validating algorithms, and
providing technical support for operational functions.
Third, promoting collaboration with universities, research institutes, and technology
firms. Such partnerships can help attract data specialists to the public sector. Flexible

68