Agentic commerce refers to software agents that shop, negotiate, and pay on your behalf, initiating and reconciling transactions with minimal human input. It raises the obvious next question. What happens when the customer is no longer involved?
In Asia, a region with many countries heavily relying on QR codes for payments, providers like Mastercard are quickly switching gears. In April 2026, Mastercard went live with authenticated agentic transactions in Singapore and Malaysia, the first wave of a rollout it plans to extend across ASEAN.
Stakeholders across Southeast Asia’s payments industry now call agentic commerce the next inflexion point, the breakthrough that comes after the QR code. The interest is not theoretical. Card networks, banks and payment gateways are building for it in the same markets at the same time.
Why is the AI Agent Payment Wave Running in Asia First?
Asia, for one, has heavy digital payment use and rails that already cross borders. The Global Payments Report 2026 found that digital wallets accounted for 77% of regional e-commerce value and 63% of Point-of-Sale value in 2025. QR code adoption continues to climb, lifting wallet volumes as it spreads.
On the other end, cash is going through a quick freefall as a share of regional spending, with the e-Conomy SEA 2025 report from Google, Temasek and Bain projecting that cash will account for only around a quarter of transaction value across the ten ASEAN economies by 2030. That leaves roughly three-quarters running on digital methods that an agent can reach.
What’s interesting is how those digital payment methods are increasingly becoming interconnected dots. The Regional Payment Connectivity Initiative, which is a central bank effort to link national instant payment systems, began with five members and now has nine signatories, with eight national QR schemes connected. This could mean that an AI agent that plugs into one of these markets, on the right payment method, can reach a payment fabric that already spans the region.
Asia-Pacific is also the fastest-growing fintech market in the world, according to BCG’s 2026 Global Fintech Report.
Why Verifiable Intent May Be the Key to AI Commerce
The harder problem in AI agent payments involves proving that the agent paid for what the customer wanted.
An ideal use case to look into is Mastercard’s agent rollout. The feature runs on Agent Pay, which wraps each AI-initiated transaction in tokenised credentials (a system that replaces card details with a single-use digital token), alongside its Agentic Tokens and Payment Passkeys, so every transaction is authenticated and auditable, end to end.
Sitting above that is Verifiable Intent, a trust layer co-developed with Google that creates a tamper-resistant record of exactly what a user authorised when an agent acted for them. It is designed to give consumers, merchants and card issuers a shared reference point if a transaction is later disputed.
Mastercard ran the initial region-wide testing with UOB, building on its first live authenticated agentic transaction in Singapore in March 2026, which Mastercard completed with DBS and UOB, in which an agent booked a ride to Changi Airport.
The company will open a regional AI Centre of Excellence in Singapore later in 2026, which it describes as its largest innovation space in Asia-Pacific.
Alongside Mastercard, Visa is building its own Intelligence Commerce framework, while PayPal has announced its acquisition of Cymbio in a bid to enhance its agentic commerce capabilities.
Who Pays When the Agent Gets It Wrong?
While authentication solves the easy half of the problem, the harder half is complex. What happens when a legitimate agent, acting inside the limits its owner has set, buys something the customer never really wanted, misreads a prompt, or locks them into a fee they did not knowingly agree to?
Mastercard’s Verifiable Intent, developed in partnership with Google, sets out to log precisely what a user authorised when an AI agent acts for them, in a form that cannot be altered after the fact.
The aim is a shared point of trust across the ecosystem, backed by cryptographic proof of authorisation that merchants, consumers and issuers can all rely on.
Mastercard is also helping shape the standards that will govern conversational AI in commerce, working alongside industry bodies to define them.
American Express has gone a step further, committing to cover mistaken purchases made by registered agents on its network and turning liability into a deliberate design choice.
Its scheme, Amex Agent Purchase Protection, works by verifying both the person and the agent acting for them, then confirming the intent behind each purchase before it stands.
Who Owns the Customer When the Agent Buys For Them?
There’s also another facet to AI agents in payments, which points out to being impartial, or the lack of it. If a customer tells an agent to “find me the best loan” or “pay this the cheapest way”, the winner could be whichever institution the agent picks, not the one the customer happens to bank with.
McKinsey, in a separate October 2025 report, estimates that AI agents could mediate 3 to 5 trillion dollars of global consumer commerce by 2030, even under moderate scenarios.
The real contest, in this case, could really be over control rather than technology. Looking into use cases from other regions, Walmart has wired its own checkout and loyalty into Google’s Gemini, so the basket and customer data stay with Walmart even when the conversation starts elsewhere.
Payment firms such as Adyen are also pushing anti-disintermediation tooling built on tokens that keep recognising the shopper behind the agent.
In Asia, the stakes are sharper still. Banks and super-apps spent the QR decade winning the customer, and the agent threatens to hand that customer to someone else.
Keeping customers loyal, at this juncture, could very well mean embedding into the agent itself rather than waiting to be picked by it.
