Agentic commerce is advancing across Asia Pacific - but not necessarily in the way many first expected.
AI agents are increasingly becoming the front door to commerce, with more people using them to plan trips, discover restaurants, and navigate everyday decisions. While agent-led transactions are starting to take shape, they are still happening in more selective use cases today.
Consumers may still be experimenting with AI assistants, but banks, merchants and payment providers are already working through a different set of questions: how an agent proves authority to act, how permissions are managed, and how transactions remain secure when the customer is no longer directly involved in every step.
More than 50 issuers and payment partners across the region are actively testing how agent-led transactions could work in practice¹. This level of participation reflects how seriously the industry is preparing for agent-led commerce.
As this new era of commerce unfolds, the foundations required to enable agent-driven payments reliably, securely, and at scale are still being built. What happens next will depend on how well the ecosystem comes together to support this new model of commerce.
Building towards an agentic future in Asia Pacific
Asia Pacific enters this transition from a position of strength. Consumers are already incorporating AI into the shopping journey, with 74% using AI-powered tools to discover, track or learn about products².
The region has spent years building digital payment habits around mobile experiences, embedded payments and integrated platforms. Consumers increasingly expect purchases to happen quickly with minimal friction, whether they are ordering food, booking transport or managing daily expenses.
As AI becomes more capable of handling parts of those journeys, expectations around speed and convenience are rising alongside it.
Digital commerce is designed around human behaviour
Today’s online commerce infrastructure is built on the assumption that a person is present throughout the transaction.
Checkout flows, authentication steps and fraud controls are designed around someone reviewing information, making decisions and confirming actions. AI agents operate differently. They rely on structured signals rather than visual cues and can move through tasks continuously without many of the pauses traditional journeys were built around.
Discovery and recommendation fit naturally into that model. Payments introduce additional requirements.
Transactions still need to be linked to a real individual. Consumers need visibility over what an agent is authorised to do on their behalf, alongside the ability to intervene when necessary. Existing systems were not originally designed for those interactions.
An ecosystem moving on multiple tracks – across merchants and issuers
Different approaches are emerging across the ecosystem. There are variations in how agents signal intent, how merchants structure product information, and how transactions are started and authorised.
The consistency consumers expect in digital payments today rests on frameworks that have evolved over decades. Agentic commerce is still at an early stage, with multiple models taking shape at once.
As adoption grows, interoperability across platforms, markets and technologies will become more important.
Trust is even more important now than ever
The core requirements of a transaction do not change, even when an agent is involved.
An agent may act on behalf of a consumer, but the same safeguards still apply. Authorisation, oversight and risk management remain essential, even when they sit behind the interaction rather than in front of it.
Those considerations take on added importance in a region that continues to face significant payment fraud challenges. As more of the shopping journey becomes automated, confidence will depend on how consistently those safeguards operate when consumers are no longer directly involved in every step of the transaction.
Visa’s network already processes about 300 billion transactions annually, drawing on a wide range of signals to assess each interaction in real time³. Extending those capabilities into agent-initiated use cases allows new models to operate within systems that consumers, merchants and financial institutions already trust.
Readiness in real-world environments
Understanding how this works in practice requires real‑world testing.
Visa Agentic Ready was launched in Asia Pacific to help issuers test and validate agent-initiated transactions within a controlled, production-grade environment. The program has achieved significant scale with more than 50 issuing partners across ten markets. Participants can explore how credential enrolment, tokenisation, authentication and transaction monitoring operate when transactions are initiated by AI agents.
Some of this work is already extending beyond simulation. In Singapore, we have piloted agent-initiated payments with DBS using real card credentials, with AI agents completing food and beverage purchases within issuer-controlled environments.These trials provide practical insight into how authentication, consent and transaction controls function when an agent initiates the transaction⁴.
Visa is also working with leading AI ecosystem players, including OpenAI, to enable agents to securely discover, initiate and complete transactions within trusted payment environments.
These developments signal how quickly the ecosystem is moving from early pilots toward practical, scalable deployment.
What it takes to scale
Agentic commerce depends on more than capable AI agents.
Identity, consent and control still need to accompany every transaction, regardless of whether it is initiated by a consumer or an agent acting on their behalf.
Asia Pacific is offering an early view of how this could work. The next phase will depend on whether these experiences can be expanded across merchants, issuers and platforms while maintaining the trust consumers expect. This will play a key role in determining how quickly agent‑led transactions become part of everyday commerce.