AI Agents Will Enter Finance Gradually: UOB Report

The United Overseas Bank (UOB) report indicates that the adoption of AI agents in finance will unfold in stages, beginning with applications that involve tightly defined mandates and clear accountability. Near-term large-scale deployments are anticipated in machine-to-machine payments, particularly for computing resources and digital services, due to their structured, repetitive nature and lower risk profile. These machine-native payments are expected to become embedded within digital platforms, operating largely in the background.

Following this initial phase, enterprise workflows such as treasury management, procurement, and reconciliation are projected to be the next wave of deployment, potentially by 2027. These areas benefit from existing governance structures, offering a practical pathway for scaling bounded autonomy within financial institutions. Consumer adoption is expected to be slower, possibly starting around 2029, focusing on low-risk activities like budgeting and cash-flow management, where permissions are explicit and easily withdrawn. Autonomous credit decisioning is seen as a more distant prospect, potentially emerging around 2031, due to complex liability and governance considerations.

However, the widespread adoption of agentic payments faces three significant, unresolved challenges. The first is liability, as current technology can identify an agent's actions but not definitively assign responsibility for financial losses. The second concerns agent identity, as financial institutions lack an accepted framework for verifying autonomous software across multiple systems and jurisdictions. Lastly, interoperability is a barrier, extending beyond technical connectivity to include liquidity, compliance, regulation, and settlement certainty across fragmented payment rails and asset classes.