Staking their bets on the agent business, payment institutions are scrambling for the "entry point"
A "entry point" competition sparked by the rise of agent-based commerce is quietly unfolding in the payment industry. Payment institutions at home and abroad have started to seize this new "entry point" one after another. Outside China, card organizations such as Visa and Mastercard are committed to integrating agent payment capabilities into a wide range of business scenarios; in China, large payment institutions including Alipay and UnionPay are also actively building a broader agent-based commercial ecosystem.
Xu Jin joked that KFC is a technology company "delayed" by its fried chicken business.
As the overall head of KFC's digital business, she led the launch of AI (Artificial Intelligence) ordering on the KFC App last year, and jointly launched Agent ordering with "Abao" under Alipay this year.
Xu Jin told reporters that the main purpose of KFC upgrading AI ordering to agent ordering is to better serve three groups of customers: nearly 500 million households, about 300 million car owners and the vast number of young people in China.
"During the ordering process, they all have their own pain points. For example, when a family orders food, some people like spicy food while others do not, making it hard to cater to all tastes; car owners cannot use their mobile phones to order while driving; young people like to try new things, but they do not know what new dishes and new set meals KFC has launched," she told reporters. KFC hopes that when these users say they "want to eat KFC", the agent on the mobile phone or the In-Vehicle Infotainment System can quickly judge their tastes and dietary preferences based on their past purchase records, and accurately recommend KFC dishes, allowing them to complete the order quickly and easily with just "one sentence".
Xu Jin revealed that if the agent ordering function achieves good results at KFC, Yum China will have greater confidence to expand this AI agent business to other catering brands such as Pizza Hut, Taco Bell, Little Sheep, Huang Ji Huang, and Lavazza Coffee.
This reinforces the judgment of Han Xinyi, CEO of Ant Group: driven by the resonance of improved AI capabilities, upgraded user demands and the formation of a closed-loop business value, agent-based commerce will see an outbreak in the next 6 to 12 months.
He believes that when agents become the "new carrier" between hundreds of millions of people and tens of millions of merchants, a new business format centered on intent is taking shape, which will bring higher user conversion rates and more considerable incremental revenue and profits.
"It is not a re-division of the existing stock cake, but incremental revenue and profits created jointly by new supply, new demand and new division of labor," said Han Xinyi.
The reporter noticed that a "entry point" competition sparked by the rise of agent-based commerce is quietly unfolding in the payment industry. Payment institutions at home and abroad have started to seize this new "entry point" one after another. Outside China, card organizations such as Visa and Mastercard are committed to integrating agent payment capabilities into a wide range of business scenarios; in China, large payment institutions including Alipay and UnionPay are also actively building a broader agent-based commercial ecosystem.
Demand Drives the Market
Up to now, Hive Box Technology has more than 400,000 sets of smart lockers, covering 236,000 communities and over 480 million users.
However, these figures bring "happy troubles" to Xu Bin, CEO of Hive Box Technology — whether the extensive business layout can make the daily active user ratio (DAU) and user penetration rate reach 5% and 10% respectively to meet the expectations of investors.
Initially, Hive Box Technology promoted its services through short video platforms and social media, hoping to improve user conversion rate and smart locker utilization rate, but the results were not satisfactory.
By chance, Xu Bin found the solution.
Xu Bin saw someone using Alipay's "Tap to Pay" feature to pick up parcels at Hive Box directly, which changed the original multi-step process of "scanning the code first, entering the password, and finally confirming the parcel pickup". This made him realize that the intelligent, fast and convenient operation process might be the key to encouraging users to use smart lockers.
Therefore, he began to popularize the "Tap to Pick Up" parcel pickup service for smart lockers.
As a large number of elderly users who like online shopping but are not familiar with complex mobile phone operation processes have got used to picking up parcels via the "Tap to Pick Up" method, the DAU of Hive Box Technology rose rapidly, increasing from 1% to about 6.5% in a short period of time.
"At that time, I did not realize the huge commercial value behind it," Xu Bin revealed. Until one time, he asked the agent on his mobile phone "where can I get my dirty shoes cleaned", and the agent directly replied: you can first put the dirty shoes into the Hive smart locker, a dedicated person will take them to the store for cleaning, and after cleaning, the shoes will be sent back to the Hive pickup locker, and he only needs to pick up the shoes and take them away.
At this moment, Xu Bin suddenly realized that in the AI era, the "home-side service" entry point for the public is changing, and a physical network is needed to support it. Smart lockers are one of the high-frequency entry points, which can connect with agents to complete a large number of people's life services and create considerable incremental revenue.
Therefore, Xu Bin decided to cover AI services in every Hive Box community within the next two years, and build a more extensive home-side service agent-based commerce system.
Ding Wei, Senior Vice President of Huangshan Tourism Development Co., Ltd., is also interested in building agent-based commerce.
Data shows that in 2025, the number of tourists visiting Huangshan Scenic Area exceeded 5 million. Ding Wei found that more than 85% of them were individual tourists, who generally faced problems such as lack of planned tour routes, inability to find outlets for scenic spot explanation services, and failure to book hotels, accommodation and catering in time.
Initially, Ding Wei spent three years integrating various data of Huangshan cultural tourism, developing a large Huangshan cultural tourism model to provide tourists with various travel guidelines. He soon found that what the vast number of individual tourists urgently need is not only travel guidelines, but also various affordable and diverse travel packages such as "scenic spot + accommodation" and "scenic spot + surrounding facilities".
Ding Wei began to build agent-based commerce for Huangshan tourism scenarios: tourists only need to tell the agent their travel time, travel preferences, accommodation and catering requirements, consumption budget, etc., and the agent can quickly formulate a travel itinerary for them, recommend suitable scenic spots, hotels and restaurants, and tourists can directly authorize the agent to place orders for hotels, scenic spot tickets and catering services.
"Up to now, our intelligent ticket purchasing service has completed a total of 3.269 million orders with a transaction volume of 420 million yuan. In the future, we will gradually expand agent-based commerce to services such as companion tour explanation and scenic spot photo taking," Ding Wei revealed.
Seize the Entry Point
Li Jun, President of Alipay Business Group, is also making every effort to seize the "new outlet" of the outbreak of agent-based commerce.
Since the beginning of this year, on the one hand, he has promoted the open test of "Abao", the agent under Alipay, which has completed intelligent adaptation of more than 10,000 services, covering eight scenarios including travel hailing, catering ordering, and cultural tourism scenic spots; on the other hand, he has promoted cross-device interconnection between "Abao" and many mobile phone and automobile manufacturers.
He told reporters that a sound agent-based commerce system cannot rely solely on AI agents, and it also needs to connect a large number of software, hardware and merchants.
"This is a win-win situation for all parties," Li Jun told reporters. For manufacturers of smart glasses, smart watches, smartphones, automobiles and other products, equipping their products with agent-based commercial functions can enable these products to respond quickly and accurately to various life service needs of users, which helps to improve user stickiness and purchase willingness.
For merchants, agent-based commerce provides a definite path to access AI, make better use of AI, and achieve more efficient and intelligent business growth with the help of AI, so as to obtain higher user conversion rate and incremental revenue.
Li Jun revealed that up to now, "Abao" under Alipay has realized cross-device collaborative service capabilities with 5 mainstream mobile phone manufacturers, its in-vehicle services cover 16 mainstream automotive brands, and it has reached designated cooperation with more than 60 automotive enterprises; emerging terminals such as smart glasses and watches are also being deployed synchronously. In addition, McDonald's, Mixue Ice City, Luckin Coffee, AutoNavi Hailing, Didi, Deppon Express, YTO Express, Hive Box and other enterprises have successively accessed "Abao" to launch agent-based commercial services.
"On the whole, the access demand of merchants is quite active, and some development demands have been scheduled to next year," Li Jun said. Payment institutions also need to provide differentiated agent-based commercial operation services according to the AI capabilities and agent operation willingness of different merchants.
For merchants who have digital accumulation but lack AI capabilities, payment institutions need to provide one-stop intelligent hosting, convert merchants' pages, commodities and service processes into Skills that can be called by agents and standardized agent units, so as to complete low-threshold AI-enabled service access.
For merchants who have completed AI-enabled service transformation and have strong willingness to operate agents, payment institutions should assist them in agent creation, skill arrangement, task execution and operation management, so as to improve their intelligent operation capabilities such as user insight, commodity recommendation and member operation.
In Li Jun's view, with the accelerated popularization of various intelligent terminals, agent-based commerce will present new development trends of full-domain coverage, scenario-based application and instant response, which will bring extensive interconnection between multiple devices and multiple agents, making entry points ubiquitous. Facing the arrival of a new round of entry point reform, Alipay must speed up its layout.
Compared with Alipay's "extensive layout", Deng Guobiao, founder and CEO of XTransfer, a B2B cross-border trade payment platform, is promoting the extensive application of AI agents in B2B cross-border trade scenarios.
On August 12, he told reporters during the 2026 Future Foreign Trade Conference that this work is not easy to carry out, because what he is facing is the B2B cross-border trade environment with low degree of digitalization.
"The contrast is quite large. On the one hand, the digitalization degree of B2B cross-border trade scenarios is very low, a large number of authenticity verification of cross-border trade background still requires manual document collection, and the data related to B2B cross-border trade such as customs, logistics and warehousing are highly scattered, making collection and verification very difficult; on the other hand, B2B cross-border trade has extremely high compliance requirements for anti-money laundering and anti-fraud," he revealed. To this end, he embedded XTransfer's self-developed B2B cross-border trade payment large model into 72 agents, covering all links including customer onboarding KYC (Know Your Customer, which mainly refers to identity verification for individuals) review, trade background authenticity verification, and continuous anti-money laundering monitoring.
He frankly told reporters that in the AI era, the business models of small and medium-sized foreign trade enterprises are also undergoing significant changes. A large number of "super foreign trade individuals" such as One Person Companies (OPCs) are rising rapidly, actively tapping into the fragmented opportunities of global B2B trade, which brings more massive and scattered B2B cross-border trade collection demands. Under this trend, how to help super foreign trade individuals complete collection efficiently is also a new development opportunity for AI agents.
This year, he often sees many super foreign trade individuals relying on their own efforts to sell various products to Africa, Southeast Asia and other international regions, but they generally face the dilemma that overseas customers lack US dollar funds and cannot make payments.
"If AI agents can build end-to-end local collection accounts and fast currency exchange services, overseas customers can pay for goods in local currency, and super foreign trade individuals can receive RMB settlement funds for withdrawal in a very short time, which will undoubtedly bring broad development space for B2B cross-border trade agent-based commerce," Deng Guobiao pointed out.
Multiple Challenges
Tang Fei is the operation director of an air travel service enterprise. Not long ago, his enterprise joined the agent-based commercial service network built by a payment institution.
Tang Fei admitted that the main reason why the enterprise decided to participate in the construction of agent-based commerce is that the marginal revenue growth effect brought by paid traffic investment continues to decline, forcing them to find new paths to maintain business growth.
"Enterprise executives have high expectations for accessing agent-based commerce. They believe that in the agent era, when machines begin to understand human needs, more human needs will be understood and met, thus bringing unexpected incremental revenue," Tang Fei said.
He soon found that whether agent-based commerce can create unexpected incremental revenue depends to a large extent on whether agents can accurately understand the intentions of the public.
Recently, after many users booked air tickets through agents, the agents automatically popped up the insurance information for accident insurance and flight delay insurance. Some users who never buy related insurance felt the experience was poor, and turned to complain to Tang Fei's enterprise asking for ticket refunds.
Tang Fei and the payment institution held urgent consultations on how to use technical means to enable agents to accurately identify which users usually are willing to buy flight delay insurance or accident insurance, and only recommend insurance services to them.
Zhou Meng, General Manager of Alipay AI Ecology Open Business Department, told reporters that whether agents can accurately understand the intentions of the public has become a major challenge that agent-based commerce urgently needs to solve.
For example, when his friend asks the agent "how to get a concert ticket", the friend not only wants to "attend the concert", but also wants to visit the surrounding scenic spots for sightseeing. At this time, whether the agent can actively recommend combined products of concert + tourism according to his past "concert + tourism" life preference can largely create incremental revenue and user conversion.
In Zhou Meng's view, to solve this challenge, the industry still needs to strengthen co-construction — all participants in agent-based commerce actively participate in the reinforcement training of agents, so that agents can more accurately identify the real intention of users' questions based on information such as users' past living habits.
He also noticed that even if agents can accurately identify and understand user intentions, there are still multiple challenges to be solved to successfully implement related services. For example, agents need to call the App or mini-program that users are used to operating to place orders, so as to effectively improve user conversion; otherwise, many users will cancel the order when facing an unfamiliar App or mini-program. In addition, to enable agents to call the service capabilities of various merchants under the conditions of trustworthiness, security and traceable data, all parties need to formulate relevant technical standards and operation rules in advance — through mechanisms such as domain-based authorization and data isolation, all parties can efficiently cooperate to complete service performance while abiding by the trust boundary.
Zhou Meng told reporters that Alipay has launched China's first AHA (Agent Hub Access) multi-agent cross-device interconnection protocol system, covering Skills Interface Standards, Agent Hub Access protocol, and X User Interface for device perception and execution, focusing on solving problems such as difficult service adaptation and unsafe collaboration across industries, entities and manufacturers, so that users can complete the whole process of cross-agent collaboration, secure performance and transaction settlement with only "one sentence of demand".
Deng Guobiao found that the biggest challenge in building agent-based commerce in B2B cross-border trade scenarios is whether AI agents can strictly control risk control.
"In the B2B cross-border trade link, the risks of enterprise account opening and payment fraud are increasing day by day," he told reporters. Last year, an enterprise opened an account at XTransfer, and the AI agent conducted in-depth retrieval of the information it provided and found that this enterprise was involved in the bitcoin mining machine business. XTransfer quickly rejected the account opening application of this enterprise; recently, a building materials company made a cross-border trade payment to another enterprise that operates 3C products such as earphones, and the AI agent issued a payment interception instruction on the grounds that the procurement relationship was unreasonable.
Deng Guobiao told reporters that the biggest concern of the vast number of B2B cross-border trade enterprises is that their accounts will be frozen if they accidentally receive suspicious payment, which will greatly affect the capital turnover of the enterprises. If AI agents can identify and intercept suspicious payments in time, they can guarantee the security of their accounts.
Deng Guobiao hopes that AI agents can "go a step further" to realize the unification of high risk control accuracy, low cost and good user experience, and completely solve this "impossible triangle" in B2B cross-border trade scenarios.
For Tang Fei, another challenge he urgently needs to solve is how to maintain the input-output ratio. After accessing the agent-based commercial service network, the daily cost of Tokens generated by agents calling services such as air tickets, train tickets and hotel reservations exceeds 1,000 yuan. This makes the enterprise worry that with the expansion of agent-based business volume, if it cannot generate sufficiently high incremental revenue, the enterprise will be unable to cover the continuously growing daily Token expenses.
Recently, Tang Fei began to learn about the specific charging model of agent-based commerce, and plans to introduce a profit-sharing mechanism to clarify the respective interests of both parties on the basis of sharing the Token cost.
Wang Yifei, head of Alipay AI Ecology Multi-device Interconnection Business, told reporters that for Alipay, the main goal at this stage is to first expand the overall scale of the agent-based commerce market. In the future, technical service