AI-driven governance in the public finance sector: lessons from high, middle, and low-income countries for Bangladesh

Abstract

Problem Bangladesh stands a critical point to modernize its public finance sector to support its next phase of economic growth, moving from LDC status. To manage the economic shifts that will occur after graduation in 2026, it is imperative to address long-standing issues with efficiency and governance (Dawar & Dutta, 2025). To deal with the economic shifts that will occur after graduation, the government must raise tax revenue and spend funds more wisely (General Economics Division (GED), 2020). Bangladesh risks falling behind other nations and losing public trust without a clear plan to integrate modern technology in its policy solution, like- AI. Findings The country has initiated some reforms. Systems like the Integrated Budget and Accounting System (iBAS++) and electronic Government Procurement (e-GP) are already in place (Bhuiyan, Rahman, & Uddin, 2020). Bangladesh isn't quite prepared for a widespread AI rollout. It lacks a clear legal framework for AI, has a shortage of skilled workers, and possesses insufficient data infrastructure to train models (UNESCO, 2025). By taking lessons from other countries, it shows that success is possible. For example, a small AI pilot project in Armenia has shown the potential for increasing efficiency in tax collection (Farooq & Vatyan, 2025) and by establishing strong legal frameworks and trust, Estonia has emerged as a leader in digital governance with its “Bürokratt”, an AI agent for better online public services (e-Estonia, 2025). Recommendations The best path for Bangladesh is a careful, phased approach using a hybrid model that combines a cautious, "sandbox" method with strong public-private partnerships. This strategy should be rolled out in three stages, starting with small, low-risk pilot projects in the first 18 months, such as using AI for fraud detection in the e-GP system and creating a taxpayer chatbot. The second phase, from months 19-36, should expand successful pilots and build local talent by partnering with universities to create research centers, like the model used in Armenia (Farooq & Vatyan, 2025). The last stage, which begins in month 37, will concentrate on completely incorporating effective AI tools into important government systems. Integration of AI will build trust, control risks, and ensure AI adoption in a way that is sustainable and advantageous for the nation's future.