Page 65 - ISC PROCEEDINGS 21.4

mechanisms grounded in advanced analytics. This section examines three representative
models - Estonia, South Korea, and France - and draws policy implications for Vietnam.
3.3.1. Estonia - interconnected data architecture and real-time monitoring
Estonia is widely recognized as one of the world’s leading digital governments.
According to the United Nations’ 2022 E-Government Development Index (EGDI), Estonia
consistently ranks among the top 10-15 countries globally. The core foundation of this
model is the decentralized data exchange platform X-Road, which enables more than 900
public and private institutions to connect and share data securely while maintaining data
ownership and control at each individual entity.
Over 99% of Estonia’s public services are delivered online; approximately 98% of
citizens possess electronic ID cards; and nearly 100% of ST transactions are processed
electronically. In the field of public finance, ST, tax, and expenditure data is integrated in
real-time, allowing for full traceability of transaction histories. The “once-only” principle,
whereby citizens and businesses provide information only once, reduces duplication and
enhances data reliability.
In terms of PIE oversight, this model facilitates anomaly detection through cross-
sectoral data matching (tax, business registration, procurement, and ST systems). A key
lesson from Estonia is that interoperable and standardized data infrastructure is a
prerequisite for deploying advanced AI applications.
3.3.2. South Korea - integrated budgeting and advanced analysis system
South Korea stands out for its integrated public state budget and accounting
management system, dBrain (Digital ST and Accounting System). This system connects the
entire ST cycle, including ST formulation, allocation, execution, accounting, and final
settlement. According to the Ministry of Economy and Finance of South Korea, dBrain
manages over USD 500 billion in public expenditure annually, covering the entire central
government and most local governments.
Centralized processing significantly reduces manual errors and shortens transaction
processing times. OECD studies have noted that the system saves hundreds of millions of
US dollars in administrative costs each year through automation and process
standardization. In monitoring public investment, dBrain enables near real-time tracking
of disbursement progress by project, ministry, and locality.
In recent years, South Korea has further integrated BDA and ML models to forecast
expenditure demand, evaluate program performance, and detect anomalous transactions.
South Korea’s approach highlights the decisive role of a comprehensive integrated ST
system as the foundation for AI applications. Advanced analytics can only be effective
when data is standardized, complete, and consistently managed across the entire system.
3.3.3. France - risk management and legal framework for AI
In France, digital transformation in Public financial governance has been
implemented through the Directorate General of Public Finances (DGFiP). This agency
oversees tax administration, state accounting, and supervision of local government state
budgets, managing databases that process hundreds of millions of transactions annually.
DGFiP has applied BDA and AI to detect tax fraud and control expenditure risks.
According to reports from the French MOF, algorithm-based risk analysis tools have
increased the detection rate of tax fraud to over 30% of in-depth audit cases, while
optimizing audit resource allocation. In public expenditure management, risk scoring
systems help prioritize transactions or projects with a higher probability of irregularities.

64