In my previous article, I examined whether Asia Pacific's payments ecosystem is ready for agentic commerce. The answer is increasingly yes, but readiness is only the beginning.
For decades, commerce has been organised around transactions. Consumers searched, evaluated and decided before completing a purchase. Increasingly, AI agents will participate in those decisions and execute them on consumers' behalf.
As intent begins to move directly into action, trust alongside identity, permissions, risk management and authentication, all critical factors in determining how quickly agentic commerce can scale within the infrastructure.
Establishing agent identity
For agentic commerce to scale, every transaction needs a clear chain of trust, grounded in verified identity and authority. Merchants need to know which consumer an agent represents and what actions it has permission to take, while agents need a reliable way to identify legitimate merchants. Establishing these verification mechanisms will be fundamental to supporting agent-initiated transactions across the ecosystem.
Visa’s work across Trusted Agent Protocol and the Agentic Directory helps address agent legitimacy and interoperability. The directory provides a verified record of legitimate agents and merchants, helping merchants recognise trusted agents and giving agents confidence that they are interacting with legitimate businesses. The aim is to create an open model to determine which agent succeeds, by providing an underlying layer of trust that allows agent-initiated transactions to take place securely.
Enabling safe and scalable agent-initiated transactions
Agentic commerce changes not just how purchases are made, but also the information needed to support them. Some areas outlined below will all play a more important role as AI agents move from assisting consumers to acting on their behalf.
Identity, permissions and context: Identity and permissions create the starting point in agentic commerce, but issuers and merchants will need more than a payment request. They will need context: whether the transaction was initiated by a consumer or an agent, what authority was granted, and whether the action fits expected behaviour. Similarly, fraud and risk models traditionally built for direct consumer interactions will need to evolve.
Large Transaction Model (LTM): Visa’s network already analyses hundreds of signals in real time. A new LTM adds another layer of intelligence, using an AI model trained on billions of transactions to further improve fraud detection across payments while increasing authorisation performance and reducing false declines.
Merchants: For merchants, it’s about making products, availability, checkout and fulfilment easier for agents to interpret and act on. Visa’s Agent Score, created with New, allows merchants to assess whether AI agents can navigate their websites, understand their content and complete tasks. The winners will not simply be those with the best consumer interface, but those that can make their commerce environments trusted, agent-readable and ready to transact.
Learning through real-world pilots
Across Asia Pacific, early use cases are testing how identity, permissions, authentication and payment controls work together in real commerce environments. These pilots matter not because they prove agentic commerce is ready to scale, but they reveal what still needs to be solved, from recognising trusted agents and enforcing consumer authority to managing transaction-level risk.
This work is already taking shape across the region. Visa is working with partners including This work is already taking shape across the region. Visa is working with partners including Animoca Brands, Ant International, DBS, Grab, LG U+, Taiwan Mobile and Tencent to explore how AI agents can discover products, make decisions and complete transactions within user-defined permissions and controls. Together, these initiatives are helping to test the infrastructure needed for secure, agent-initiated commerce and demonstrating that scaling agentic commerce will require coordinated progress across issuers, AI platforms, merchants, digital ecosystems and payments networks.
Our collaboration with OpenAI offers another early signal of how the ecosystem may evolve. AI platforms may become the interface where intent is expressed, while payment networks provide the identity, tokenisation, authorisation and risk-management capabilities needed to enable transactions safely.
The opportunity is especially relevant in Asia Pacific, where consumer behaviour is moving faster than the commerce infrastructure around it. While AI-driven discovery, decision-making and transactions are beginning to move from concept to reality, the pace of adoption will ultimately depend on how quickly the broader ecosystem can evolve to support them.
Looking ahead
AI will continue to evolve, and that is a given. The real challenge is not intelligence. It is trust.
Trust is what connects consumers, agents, merchants and financial institutions, turning decisions into transactions. This is where Visa is taking an active lead: building the networks, partnerships and capabilities that make agent-initiated commerce possible.