Banks in India and Singapore Deploy Autonomous AI for Core Operations
Financial institutions are moving agentic AI into production for mission-critical tasks. Seven public sector banks in India have rolled out 32 generative AI use cases, including fraud detection and customer service automation. In a significant step toward financial autonomy, Singapore’s DBS Bank is piloting a system where AI agents are authorized to make payments on behalf of customers under strict risk controls.
- The initiative by seven public sector banks in India is part of the EASE 8.0 program, which focuses on modernizing state-run lenders through technology and risk-management upgrades. The 32 generative AI use cases are primarily aimed at improving credit appraisal by analyzing a borrower's business overview, financial history, and repayment track record to support faster, more informed decisions. - The DBS Bank pilot is a collaboration with Visa, utilizing its Intelligent Commerce (VIC) framework. This architecture uses tokenized credentials and keeps the bank in control of the transaction flow, requiring issuer approval to verify the user's identity and spending permissions before a payment is executed by the AI agent. - Agentic AI architectures in banking are moving beyond single-task models to multi-agent systems that orchestrate complex workflows. These systems typically consist of a perception layer for data intake, a cognition layer with LLMs for reasoning, and API connectivity to core banking systems for executing actions. - Singapore’s approach to AI governance is guided by its Model AI Governance Framework, which is voluntary and emphasizes principles like accountability, transparency, and human oversight. The government also launched the AI Verify Foundation to develop testing tools for organizations to validate the performance of their AI systems against these principles. - India's AI governance is taking shape through a techno-legal framework that balances innovation with risk mitigation. In August 2025, the Reserve Bank of India's "FREE-AI Committee" recommended seven principles to guide AI development in the financial sector, focusing on trust, fairness, and human-centricity. - A key challenge in deploying agentic AI across multiple jurisdictions is data localization. For a bank operating in the EU, India, and the US, transaction data cannot leave the originating geography, requiring a distributed architecture where AI agents operate locally but are managed by a central control plane for oversight and reporting. - Enterprise adoption of financial AI is maturing from isolated use cases to automating end-to-end workflows. Common architectural patterns combine document AI platforms like ABBYY for data extraction with RPA tools like UiPath or orchestration platforms like IBM for managing multi-step processes across different systems. - Risk management frameworks for AI in finance are evolving to address model explainability, data drift, and integration security. Regulators expect financial institutions to implement robust governance that includes continuous monitoring of AI models, comprehensive documentation, and clear human oversight with the ability to intervene in or override AI-generated decisions.
