Open the artificial intelligence interface from Alipay’s home screen, and the familiar grid of mini programs gives way to a chat box. A user can ask “Ah Bao” to “find milk tea coupons nearby.” The app matches offers from nearby stores, then handles the coupon, order, and payment in a single exchange.
The feature follows what is arguably the largest redesign in Alipay’s 22-year history.
Earlier this month, Alipay opened Ah Bao, its AI-enhanced interface, to public beta. Around the same time, WeChat Pay introduced an “AI card” within the WorkBuddy agent.
Since the start of 2026, major platforms and payment providers have unveiled their approaches to AI in quick succession. Agentic AI, systems designed to carry out tasks on a user’s behalf, is beginning to enter payments.
But as payment apps adopt AI, they are also encountering behavior unlike the utilitarian habits they were built to encourage.
After the AI version of Alipay entered closed beta, a company representative told 36Kr that one development surprised the team. Many users opened Ah Bao not to complete a task, but simply to “take it for a spin.” They chatted with it, asked unusual questions, and tested its responses.
“Ah Bao’s human touch has made users more interested in interacting with it,” the representative said. “Many people come just to play with Ah Bao, which surprised us as well. That would have been highly unlikely with the traditional version of Alipay.”
The change is revealing. For the past decade, payment apps have tried to become efficient, frictionless, and easy to leave as soon as a transaction is complete. The less time users spent inside them, the better the experience was thought to be.
Alipay, like other payment tools built to support particular transactions, has faced natural limits on retention. In recent years, value-oriented e-commerce, content-led commerce, local retail, and other formats have drawn users into new channels, further fragmenting traffic once concentrated on marketplace platforms. Even Alipay has frequently relied on digital cash gifts, discounts, and subsidies to encourage activity.
AI may reverse that pattern by moving payment tools from the final stage of a purchase toward the beginning.
Previously, someone buying milk tea might open a food delivery platform, place an order, and select Alipay only at checkout. Payment was the last step.
Now a user can describe what they want in Ah Bao’s chat box and complete product selection, coupon collection, ordering, and payment without leaving the conversation. Ah Bao is becoming an entry point for purchase decisions, not merely a tool for settling them.
Payment apps are beginning to acquire the sustained attention usually associated with services where a transaction originates.
Across the industry, AI is changing how users encounter payment services. The prevailing view is not that AI will replace existing payment flows. Rather, it will add a layer of decision-making at the application level, influencing what users see, choose, and buy before payment begins.
Alipay has taken the broadest approach by redesigning the product itself.
“Alipay’s AI strategy is a choice made for the future of the agent economy,” an Alipay representative said. “We believe the relationship between people and services will be restructured. The app-based intermediary of the past will be replaced by agents, and a new architecture is required to support that change.”
According to information obtained by 36Kr, Alipay began exploring AI in the second half of 2023. It started with an on-device voice assistant, then introduced specialized agents for government services, transportation, employment, and other fields. Only after gathering sufficient operating data did it begin rebuilding the app at the client level.
The redesigned version of Alipay takes tens of thousands of services once buried under layers of menus and gathers them behind a single chat box.
It now supports 72 frequently used tasks, including reviewing account activity, paying bills, accessing government services, and arranging transportation.
Alipay has also opened the system to outside providers. The company recently introduced an AI platform that allows eligible third-party services to connect with Ah Bao through the Model Context Protocol, or MCP, and other skill protocols.
WeChat Pay has chosen a different approach.
Rather than building a centralized AI interface, it introduced an AI card as a payment function that third-party agents can place inside their own service flows.
When a user expresses an intention to buy something within an agent, the AI selects products and initiates the transaction. The card then handles the payment.
A WeChat Pay representative told 36Kr that the card has a narrow purpose: to make agent-based payments secure and controllable, then allow third-party platforms to use that function.
“The agent in which a user develops a purchasing need, how the agent interprets that need, which products it recommends, and how the interaction is designed are all part of the value provided by each agent platform,” the representative said.
“WeChat Pay only steps in at the final payment stage. Within the scope authorized by the user, it provides a system in which funds are reserved for designated purposes and every transaction requires confirmation.”
Alipay is rebuilding its product around an AI interface. WeChat Pay is offering a standardized payment component. The two products are not directly comparable.
Viewed through the changes taking place in payments, however, both point in the same direction. Payment is moving beyond a supporting role and closer to the moment when a user decides what to buy.
The shift from passive to proactive service is visible across the industry, though providers are pursuing different routes. Those choices reflect the data, services, and commercial relationships available to each company.
Over Alipay’s 22 years of operation, its service and transaction data have become closely connected. According to the company, its ecosystem includes more than one million mini programs and over 8,000 services.
They range from wealth management and insurance to utility payments, government services, and transportation bookings. Many already operate inside Alipay, giving its AI a large supply of services that it can coordinate directly.
Alipay’s offline network provides another advantage.
On July 8, the company announced an AI upgrade to its “Tap!” feature. Alipay said its 30 million offline touchpoints would become part of an AI-assisted commercial network, while devices used by more than one million merchants would function as “device agents.”
With a tap, users can access coupons, membership services, marketing campaigns, and other offerings coordinated by Xiaoyu, Alipay’s merchant-facing agent. The system extends Alipay’s AI interface into physical retail locations.
Ah Bao handles online interactions, while “Tap!” serves offline ones. Alipay’s approach therefore has two parts: a conversational interface for digital services and a network of physical touchpoints for merchants. Both operate on the same open AI foundation.
WeChat Pay’s advantage is the breadth of the WeChat ecosystem.
WeChat contains a vast number of accounts, mini programs, channels, and third-party agents. Consumer demand is spread across many settings rather than concentrated in a single service interface.
For WeChat Pay, the more practical approach is to make payment a standardized component, open it to external developers, and allow different agents to connect to it securely.
In June, JD.com released its A2P2 autonomous-agent payment protocol. Drawing on its e-commerce and supply chain operations, the protocol provides different implementation levels for different situations.
Its payment functions cover consumer voice payments and automated business-to-business supply chain settlements, allowing JD.com to move relatively quickly in enterprise settings.
The routes chosen by Alipay, WeChat Pay, and JD.com reflect the resources each already possesses.
Companies with their own service ecosystems are more likely to build centralized interfaces. Those operating across dispersed services tend to emphasize connectivity. Companies with extensive supply chain operations are more inclined to begin with business applications.
Together, these approaches show how AI is beginning to alter the payments industry.
Whatever route a provider chooses, the same question remains: will users trust AI with their money?
The answer may depend less on technical capability than on whether users feel they remain in control.
The industry’s early systems share one principle. The user keeps final authority over the transaction.
WeChat Pay separates the AI card’s funds and requires confirmation for every payment. The card sits within the user’s WeChat Pay balance. Users transfer money into it themselves, keeping the funds separate from their main accounts. For each transaction, the AI selects a product, places the order, and initiates payment. Before money is deducted, the user must confirm the purchase and enter a password. Funds can be moved into or out of the card at any time.
“The prerequisite for users to feel comfortable handing money to AI is that they always know, and always control, how much the AI can spend and what it can spend the money on,” the WeChat Pay representative said.
Alipay has built its safeguards around identity verification, traceability, and spending controls.
The system includes a proprietary identity-verification method, end-to-end transaction records, and limits on authorized spending. Alipay says these measures are intended to contain unauthorized transactions and misuse when payments move across platforms. Every action taken by an agent, including each payment, can be traced.
For now, restraint appears to be the industry’s common position.
KrASIA features translated and adapted content that was originally published by 36Kr. This article was written by Wang Hanyu for 36Kr.