Agentic commerce Singapore: 2026 AI payments guide

Key takeaways:

• Singapore is APAC's live testing ground for agentic commerce. Mastercard Agent Pay and Visa are already transacting with DBS and UOB, making us the first market in Southeast Asia with authenticated agentic transactions.

• APAC is set to drive around two thirds of the world's new retail sales over the next five years, with agentic AI adoption among consumer businesses expected to jump from 29% to 76% within two years¹.

• For Singapore merchants, agent-readiness isn't a single feature: it spans product discovery, payment infrastructure, fraud controls, and post-checkout liability. Airwallex's multi-currency Business Accounts and programmable payment APIs are built for exactly this environment.

Agentic commerce in Singapore is no longer something to prepare for; it's already here. AI agents are executing real purchases on live payment rails, and the first authenticated agentic transaction in Southeast Asia took place in Singapore in March 2026.

For merchants, the question is no longer whether AI agents will become buyers, but whether their business is ready when they do.

This guide explains what's already happening in Singapore, the regulatory and payment infrastructure supporting it, and the practical steps merchants can take to stay competitive in an agent-driven commerce landscape.

Singapore as APAC's agentic commerce hub

Singapore is leading the shift to agentic commerce. A combination of infrastructure, regulation, and institutional investment has made our country the natural home for APAC's agentic commerce buildout.

Why Singapore is first

Several factors set Singapore apart from the rest of the region:

• The Monetary Authority of Singapore's (MAS) clear regulatory posture on AI in financial transactions gives banks and networks the certainty they need to deploy.

• PayNow and FAST real-time payment rails provide the settlement infrastructure underneath.

• Singapore's density of regional headquarters means that what launches here tends to roll out across APAC next.

Mastercard's AI Centre of Excellence, based in Singapore, is a direct signal of where the global networks are placing their bets.

The ASEAN rollout

Singapore and Malaysia are the first two ASEAN markets with live agentic payment deployments. For merchants with cross-border operations, that matters now. Agents transacting on behalf of buyers in either market will expect payment infrastructure that handles multi-currency settlement without friction.

Since Singapore is ASEAN's agentic commerce launchpad, your payment infrastructure needs to be ready for what's coming. Airwallex Payments gives you the multi-currency foundation to accept transactions across SG, MY, and beyond. Learn more about Airwallex Payments or sign up now.

What's already live: the first agentic transactions in Singapore

Here's what has already been deployed in Singapore, and what each one tells merchants:

1. Mastercard Agent Pay (DBS and UOB)

Southeast Asia's first authenticated agentic transaction took place in March 2026: a ride booking to Changi Airport via hoppa, processed on Mastercard Agent Pay. Both DBS and UOB are now live on the network.

The authentication layer runs on Payment Passkeys, a credential standard that lets an AI agent verify a transaction without exposing raw card details. The agent presents a bound credential; the network validates it; the transaction clears. No human click required.

2. Visa pilot (DBS)

DBS is also live on a separate Visa agentic pilot, covering food and beverage purchases.

Where Mastercard Agent Pay uses tokenised credentials, Visa's pilot runs on real card credentials: a meaningful technical distinction that affects how merchants handle reconciliation and disputes downstream.

3. Visa Agentic Ready (13 Singapore issuers)

On 30 April 2026, Visa expanded beyond the DBS pilot with the launch of Agentic Ready in Singapore. The programme now includes DBS, HSBC Singapore, OCBC, Standard Chartered, GXS, Maybank Singapore, and seven other local issuers.

This is the clearest signal yet that agentic payments are moving from isolated pilots into production infrastructure across the issuer ecosystem.

4. HSBC / SourceSage / FortyTwo: B2B agentic transaction

In May 2026, HSBC, SourceSage, and Singapore eCommerce merchant FortyTwo completed an end-to-end B2B agentic transaction on Mastercard Agent Pay.

Unlike the Changi ride booking, this was a procurement workflow: a supplier purchasing use case that most Singapore merchants will recognise. Agentic commerce is no longer just a consumer story.

What these deployments tell merchants

Agents are transacting on live rails today across consumer and B2B contexts. If your checkout isn't built to receive agent-initiated payments, you're already outside the transaction flow for a growing share of buyers.

MAS, IMDA, and the legal rules Singapore merchants need to know

Singapore's regulatory environment for agentic payments is structured but not restrictive. Three frameworks shape what merchants are responsible for, and where their liability sits.

MAS and the agentic payments environment

MAS has not created a separate regulatory category for AI-initiated payments.

Instead, existing frameworks — consumer protection, anti-money laundering (AML), and know your customer (KYC) requirements — apply to agent-initiated transactions in the same way they apply to any other payment. The agent is treated as an extension of the account holder, not a separate regulated entity.

MAS's BLOOM initiative (Borderless, Liquid, Open, Online, Multi-currency) sits underneath this as the settlement infrastructure layer. It's the framework guiding Singapore's push toward real-time, multi-currency digital payments, and it's what makes the PayNow and FAST rails capable of supporting agentic transaction volumes.

IMDA Model AI Governance Framework

The Infocomm Media Development Authority's (IMDA) Model AI Governance Framework, published in January 2026, is the world's first structured guide for responsible agentic AI deployment. 

For merchants, the key principle is practical: use deterministic, rule-based controls to define what an agent can and cannot do, rather than relying on prompt-level instructions alone.

The framework also sets out human checkpoint requirements at high-value or high-risk transaction thresholds. If an agent is purchasing above a defined ceiling, a human confirmation step may be required: merchants need to build for this in their checkout logic.

The Electronic Transactions Act and AI-initiated contracts

Under Section 11 of Singapore's Electronic Transactions Act (ETA), contracts formed by automated systems are legally valid. An agent completing a purchase on a user's behalf creates a binding agreement.

The operative legal principle is instrumentality: AI is treated as an extension of the human operator, not an independent party. The Court of Appeal applied this logic in Quoine v B2C2, where programmer intent was used to determine whether an algorithmic trading error constituted a valid contract. The same reasoning applies to agentic commerce disputes.

Merchants should maintain robust audit trails because responsibility for disputes and liability depends on the applicable legal and payment-network frameworks.

How agentic payments work: identity, authorisation, and settlement

Before a merchant can accept an agentic payment, three things need to be true: the agent's identity is verified, the transaction falls within the user's pre-authorised rules, and the payment clears on infrastructure that understands both. Here's how each layer works.

The Know Your Agent (KYA) standard

Know Your Agent (KYA) is the emerging cross-industry standard for verifying that an AI agent is who it claims to be, and that it's acting within the scope its user authorised.

Without KYA, merchants have no reliable way to distinguish a legitimate buying agent from a malicious bot running automated checkout attacks at machine speed.

Leading KYA protocols in use

Four protocols are active in the market today, each with a different technical approach:

1. Mastercard Agentic Tokens embed agent identity directly into network-level token metadata — the approach underpinning the DBS and UOB deployments in Singapore

2. Visa TAP (Trusted Agent Protocol) uses RFC 9421 HTTP request signatures to bind agent identity to canonical request headers — the method used in the DBS Visa pilot

3. Skyfire KYAPay uses claims-based JSON Web Tokens (JWTs) for OAuth-friendly user delegation, letting operators define and verify the scope of what an agent can do

4. Google AP2 Mandates embed cryptographic user-signed mandates directly inside the request payload, so the spending authority travels with the transaction

Human-to-agent delegation: scoped spending and mandates

Users don't hand agents a blank cheque. They set spending ceilings, category restrictions, and domain-specific rules in advance, also known as a “mandate”. When an agent initiates a transaction, that mandate data travels with the payment request.

Merchants need to receive and respect mandate data, not just process the transaction. An agent operating outside its mandate is a liability event for everyone in the chain.

What this means for Singapore merchants

Agent-readiness isn't a single switch to flip. It spans how your inventory is structured, how your payment stack is built, how your fraud controls are tuned, and how you handle disputes after the fact. Here's what to prioritise:

1. Make your inventory visible to agents

Agents don't browse your storefront; they query structured data feeds. If your product catalogue isn't machine-readable, you're invisible to the agent before the transaction even starts.

How do you do this? Here are a few steps to take:

• Audit your catalogue metadata for completeness: price, availability, delivery conditions, and contextual attributes all need to be present and accurate.

• Deploy schema.org markup and Model Context Protocol (MCP) server endpoints so AI agents can parse and query your inventory directly.

• Make sure your feed infrastructure can support more frequent automated requests from AI agents.

2. Upgrade your payment stack for agent-initiated transactions

Many existing checkout flows such as forms, redirects, and CAPTCHAs aren't navigable by agents. They're built for humans, and agents will either fail at them or route around them to a merchant who doesn't have them.

You need infrastructure that accepts tokenised credentials, processes mandate data, and authenticates across multiple agentic protocols. Agents transacting across borders will initiate payments in the buyer's local currency, so multi-currency capability is a baseline requirement.

Airwallex's Business Accounts let you collect funds in 70+ countries, with programmable payment APIs built for exactly this kind of automated transaction flow.

3. Rethink fraud controls for machine-speed commerce

Agentic traffic looks like bot traffic: high-frequency inventory checks, rapid sequential requests, automated checkout flows. Risk systems calibrated to block bots will block legitimate agents too, and that means lost revenue.

The answer is differentiation, not blanket blocking. KYA protocols give you the identity layer to tell the difference. Work with payment partners who integrate KYA natively, so your fraud controls can verify agent credentials rather than just pattern-matching on behaviour.

4. Build for post-checkout accountability

When an agent acts outside a user's expectations, disputes follow. Incomplete transaction records can make chargeback disputes more difficult to resolve.

Agentic transaction data needs to capture mandate context, authorisation records, and agent identity at the point of sale. Post-checkout accountability is directly tied to your liability exposure.

5. Build category authority to avoid disintermediation

When an agent selects a merchant on a user's behalf, the decision isn't based on your homepage design or brand recall: it's based on data quality, catalogue completeness, and established preference signals. Merchants with poor structured data may be less likely to be selected by AI agents.

L'Oréal's Noli platform is a useful model here: it built a direct, data-rich relationship with AI agents at the category level, rather than waiting to be discovered through a generic product feed.

For Singapore merchants, establishing an API-accessible catalogue early puts you in the agent's default consideration set before your competitors get there.

Why Singapore merchants choose Airwallex for agentic commerce

Most payment infrastructure wasn't built for agentic transactions. Forms, redirects, and manual authentication steps assume a human is present, and agents can't navigate them.

In contrast, Airwallex is built around programmable, API-first payment infrastructure:

• Our Global Accounts let you hold in 20+ currencies from a single account, which is critical when agents are transacting across SG, MY, and beyond on your buyers' behalf.

• Our Payments infrastructure supports multi-currency collection across 130+ currencies and 180+ countries, with local rails in 120+ markets so cross-border agentic transactions clear without unnecessary friction or conversion costs.

Frequently asked questions (FAQs)

What is the difference between a shopping bot and an agentic commerce system?

A shopping bot helps with discovery: it searches, compares, and surfaces options, but you still confirm and complete the purchase. An agentic commerce system executes the full transaction autonomously, from product discovery through to payment, within rules and spending limits you define in advance. The key distinction is who clicks "pay": with agentic systems, no one does.

Which banks in Singapore currently support agentic payments?

DBS and UOB are live on Mastercard Agent Pay, and DBS also participates in Visa's agentic pilot covering food and beverage purchases. On 30 April 2026, Visa launched its Agentic Ready programme in Singapore with 13 issuers, including OCBC, HSBC Singapore, Standard Chartered, GXS, and Maybank Singapore. The issuer base is expanding quickly.

How does Singapore's Electronic Transactions Act govern AI-made purchases?

Section 11 of the Electronic Transactions Act (ETA) confirms that contracts formed by automated systems are legally valid. AI has no independent legal status under Singapore common law: the operator who deploys the agent generally remains responsible for its actions under applicable legal frameworks. The Court of Appeal's ruling in Quoine v B2C2 established that programmer intent is the determining factor in resolving disputes over automated transactions.

Can an AI agent hold or open a business bank account in Singapore?

No. All financial accounts must be held by a verified human or registered legal entity under MAS regulations. An AI agent can be authorised to transact on behalf of an account holder, but it cannot own or open an account in its own right.

How do KYA protocols protect Singapore merchants from automated fraud?

Know Your Agent (KYA) protocols use cryptographic tokens and request signatures to verify that an agent is legitimately authorised and acting within its defined scope. This lets merchants differentiate authorised buying agents from malicious bots without applying blanket blocks that would also stop legitimate agentic transactions.

What payment infrastructure do merchants need to accept agentic transactions?

You need checkout flows that don't require human navigation, infrastructure that accepts tokenised credentials and processes mandate data, and multi-currency settlement capability for cross-border agentic transactions. Programmatic APIs that work across multiple agentic protocols are the baseline.

Sources

1. deloitte.com/ap/en/about/press-room/apac-set-to-lead-the-agentic-future-of-commerce.html

This publication does not constitute legal, tax, or professional advice from Airwallex, nor does it substitute seeking such advice, and makes no express or implied representations / warranties / guarantees regarding content accuracy, completeness, or currency. If you would like to request an update, feel free to contact us at [[email protected]]. Airwallex (Singapore) Pte. Ltd. (201626561Z) is licensed as a Major Payment Institution and regulated by the Monetary Authority of Singapore.