DBS and Visa Trial AI Agents for Automated Customer Payments

Artificial intelligence is beginning to move beyond its advisory role in finance, with DBS Bank launching a pilot that enables AI agents to execute payments on behalf of customers. In collaboration with Visa, the Singapore-based lender is testing the Visa Intelligent Commerce framework, which facilitates transactions directly initiated by AI software instead of human users.

The trial allows digital agents to search for products, select items, and complete purchases using payment credentials issued by DBS or its subsidiary POSB. Early efforts have already processed real transactions, such as food and beverage purchases, according to reports from Asian Banking & Finance and Fintech Futures.

AI Agents in Payments: Moving Beyond Advice

This pilot illustrates the financial sector's readiness for what is being called "agent-driven commerce." Under this concept, AI tools act according to rules set by both customers and the issuing bank. Visa's system keeps the bank at the core of approvals: payment details are tokenised, and every transaction must pass through a bank-controlled approval process, which verifies identity and enforces spending limits. In effect, financial institutions retain final control over whether an AI agent's transaction proceeds, ensuring regulatory compliance and customer consent.

Banks' exploration of "AI agents" reflects a wider shift from using AI as a customer-facing tool to integrating it deeper into payment infrastructure. Previously, AI in banking primarily supported functions such as productivity assistance and data analysis. The new approach sees AI becoming an active participant in transaction execution—requiring new safeguards and oversight.

Initial Use Cases and Controls

Current applications of agent-based transactions focus on routine, low-risk purchases, such as recurring grocery orders, subscription renewals, travel bookings, and household re-stocking. These transactions are typically guided by parameters set by the user, including budgets and brand preferences. DBS and Visa are planning a phased expansion of the pilot to cover broader online shopping and travel payments.

This technology presents both opportunities and challenges for the banking industry. On one hand, banks that can securely facilitate AI-initiated payments may enhance their role as digital commerce enablers, acting as the consent and control layer for automated transactions. On the other hand, new issues regarding liability and dispute resolution are emerging, particularly if customers contest purchases made by AI agents on their behalf.

Security and governance are expected to influence the pace of adoption. Analysts note that while customers may readily accept AI-generated suggestions, they may hesitate to trust AI with direct financial decisions. Maintaining strong controls—such as keeping transaction approval logic within the issuing bank—may help reassure customers and encourage adoption.

Banking Sector Integrates AI Deeper in Operations

The introduction of agent-based payments is part of a broader trend in enterprise AI deployment. In the past year, banks have moved beyond testing chatbots and internal assistance tools, implementing AI into functions that influence revenue, operations, and customer interactions. This includes applications in fraud detection, credit analysis, and automated customer service. Allowing AI to trigger payments represents a significant step toward full automation.

DBS has been a leader in digital transformation, investing substantially in data analytics and AI to optimise operations and tailor services. The current pilot builds on these foundations, aiming to determine how AI might ultimately streamline and secure the payment process at scale.

The normalization of agent-based financial transactions will depend heavily on customer trust and the clarity of limits placed on AI agents by banks. Industry experts predict adoption will likely begin with low-value, frequent transactions before expanding to more complex purchasing scenarios.

For further reading, see the original source at artificialintelligence-news.com.