E-commerce is present process a profound transformation, triggered by synthetic intelligence (AI) brokers and the transfer from “machine-to-machine” (M2M) funds to actually autonomous “AI-to-AI” funds, says Sherman Jiang, a senior product government specialised in shopper finance, fintech, and banking.

This shift requires an entire rethinking of cost programs and introduces new challenges for product leaders to deal with

In a latest put up revealed on September 12, Jiang appears at the way forward for e-commerce and the potential impression of agentic AI on the area, and descriptions the hurdles forward.

In response to Jiang, in contrast to their older, rule-based M2M funds, agentic funds contain reasoning, studying, and performing with a objective. These programs transcend merely reordering a product when a sensor hits a preset threshold, and are as an alternative able to autonomously detecting {that a} consumer is operating low on a product, looking for the best-rated choices from native producers, negotiating a reduction based mostly on flash gross sales, putting the order, and paying for it.

For product managers, this marks an unprecedented shift the place the “buyer” is not a single particular person anymore, however moderately an clever, dynamic AI agent that wants new instruments, new rails, and new safeguards. This requires product executives to construct higher wallets that may delegate funds securely, enabling commerce to turn into genuinely autonomous, he says.

Fraud and legal responsibility challenges

Although the shift to agentic commerce presents large alternatives to enhance buyer assist, drive better effectivity, and supply tailor-made buyer experiences, these adjustments additionally introduce new challenges, Jiang warns.

As a result of AI brokers function at machine velocity and scale, their exercise might resemble legal habits beneath immediately’s fraud detection fashions. This may require an entire overhaul of fraud prevention methods, transferring away from the know-your-customer (KYC) fashions and in the direction of so-called “know-your-agent” frameworks.

The important thing right here can be to construct a belief infrastructure that’s able to distinguishing a “good” AI from a “dangerous” one, a system which is able to probably depend on cryptographic tokens to show id and authority.

For instance, a consumer might show their id with a Person ID Token. This token would act as a digital identification card that confirms the particular person is actual and verified. Equally, an AI agent performing on behalf of a consumer could possibly be issued an Agent ID Token, which might work like a passport for the AI and set up its id within the system. Lastly, a Delegation Token might join the 2, and clearly outline what the AI is allowed to do. Collectively, these three tokens would type a series of belief between the human, the AI agent, and the connection between them.

In addition to fraud, the shift in the direction of agentic AI additionally raises new questions on legal responsibility and dispute decision, specifically when an AI agent makes a purchase order the consumer didn’t intend.

Most conventional authorized frameworks rely upon the concept that somebody acted deliberately or negligently. Nonetheless, as a result of AI brokers make purchases which can be technically approved however which can go towards the consumer’s precise intent, it produces a legal responsibility gray space.

Compounding that is the so-called “agentic loyalty downside”, Jiang provides, the place AI brokers might prioritize the pursuits of the platform that deployed it moderately the consumer’s greatest pursuits. For instance, an agent might select a dearer merchandise from a companion firm even when a less expensive, extra appropriate possibility exists elsewhere.

Incompatibility and infrastructure gaps

Lastly, the rise of agentic commerce additionally exposes limitations in legacy monetary programs. These programs, which embody bank cards, subscriptions, and batch settlements, are constructed for human-centric, high-value, low-frequency transactions. This makes them ill-fitted for high-frequency, low-value transactions that AI brokers might have to make, equivalent to paying for particular person API calls, information, or computing sources on a “pay-per-use” foundation, usually for fractions of a cent.

To deal with this, crypto-native expertise gamers are coming into the area with blockchain-based micropayment programs as the inspiration for this new financial system. Coinbase, for instance, launched in Could 2025 x402, a cost protocol that allows immediate stablecoin funds immediately over HTTP. The answer is designed to let APIs, apps, and AI brokers to transact seamlessly.

Past funds, blockchain-based infrastructure might additionally provide the core primitives that conventional programs lack, Jiang argues. This features a safe and immutable ledger, verifiable identities, and sensible contracts for autonomous worth switch.

AI brokers see speedy adoption

AI brokers are quickly coming into digital commerce, pushed by rising buyer demand. A July 2025 US shopper survey by Boston Consulting Group (BCG) discovered that 81% of customers anticipate to buy utilizing agentic AI.

BCG estimates that within the coming years, greater than US$1 trillion in spending, representing about 50% of complete e-commerce expenditure immediately, could possibly be agent assisted. Early adoption will probably be concentrated in routine and repeat purchases, equivalent to family provides, restaurant orders, private care, and nutritional vitamins/dietary supplements, moderately than high-ticket or high-emotion purchases equivalent to luxurious items or medical gadgets.

Main gamers, together with Google, PayPal, Visa, and Mastercard, are already racing to seize the agentic AI shift.

Fee Webinar

As APAC’s cost panorama continues to evolve, Fintech Information Singapore can be internet hosting a webinar on September 24, 2025 at 3:00PM to discover the regional traits shaping the way forward for the trade. The session will dive into the potential of Click on to Pay, in addition to the newest improvements in passwordless authentication, and can characteristic prime trade executives and specialists from main organizations, together with Worldpay, Visa, and Thales.

Featured picture by tete_escape on Freepik