Is the intercommunication between "Xiaomei" and Yuanbao setting an example for the WeChat intelligent agent?

After Meituan released its financial report for the first quarter of 2026, the outside world focused on the better - than - expected "loss reduction" figures. However, another piece of news revealed by Wang Xing during the earnings conference call didn't receive the attention it deserved.

He said that Meituan's AI assistant "Xiaomei" will be connected with Tencent's Yuanbao. When users express their local life needs in Yuanbao, the system will automatically trigger "Agent to Agent" communication with Xiaomei, seamlessly activating Meituan's food delivery and other services.

This "publicized in advance" cooperation not only points out Meituan's strategic direction of "strengthening infrastructure and connecting with external AI entrances", but also sets an example for the next - step in - depth cooperation with WeChat's intelligent agent.

Coincidentally, on the day after Meituan announced its partnership with Tencent's Yuanbao, some media reported that Doubao had recently completed the connection with Douyin's local life service business. With 340 million monthly active users, this national - level AI entrance is becoming a sharp blade stabbing at Meituan's core business.

On the same day, the Financial Times also reported that WeChat's intelligent agent had completed the prototype test and would start the compliance approval process required for public launch as early as this month. This means that Tencent is accelerating the connection work with millions of mini - programs. Since WeChat's intelligent agent involves 1.4 billion users, the preparations before the launch are bound to be extremely cautious. The interconnection between Yuanbao and Xiaomei can serve as a preliminary "stress test".

Under the pressure of profitability, domestic AI assistants are accelerating the transition from the Chatbot era and competing to become entrances for real - world transactions and life services. This also means that Meituan's self - positioning and its relationships with Tencent, ByteDance, and Alibaba are changing rapidly. Although the "connection" between Xiaomei and Yuanbao seems like a small move, it may well be the prelude to a new round of business alliances.

Meituan's moat is being bypassed by Doubao and Qianwen

Meituan was able to significantly "reduce losses" in the first quarter mainly because its advantage in the mid - to high - end dinner market remained almost unshaken. This is the real demand that remained after the food delivery subsidy war and is Meituan's most solid moat.

During the conference call, Wang Xing said, "Meituan's real moat lies in its structural data advantages, namely, comprehensive local services covering all categories, verified real - time information, real user reviews, and a powerful fulfillment infrastructure. By integrating these structural advantages with AI, we will provide users with the best AI - powered local life service experience."

The problem is how to integrate Meituan's local life infrastructure with AI and whether it can resist the "walled gardens" powered by AI from Alibaba and ByteDance on its own.

Although Meituan's AI investment covers areas such as large - scale models, embodied intelligence, and computing chips, and includes well - known companies like Yuezhianmian, Yushu Technology, and Yinhe Tongyong, most of these are long - term ecological positioning moves and do not directly resonate with its main business.

Currently, Meituan has given up the ambition of building an independent AI entrance. Xiaomei and Xiaotuan have also retreated to the app and are placed in the core position of the bottom navigation bar. This approach has achieved phased results: during the Spring Festival holiday, "Xiaotuan" served over 100 million users; during the May Day holiday, the query volume of "Xiaotuan" increased by nearly 90% compared with the Spring Festival, and the average message read - through rate reached 61%. On the merchant side, more than 3.4 million merchants are using Meituan's AI business assistant.

However, Xiaomei and Xiaotuan are mainly tools to improve the efficiency of using the Meituan app, rather than independent traffic entrances. This limitation is becoming increasingly dangerous in the current situation where AI entrances are reshaping the rules of traffic distribution.

Although Meituan still has no rivals in terms of fulfillment efficiency, Doubao, Qianwen, etc. are cultivating users' consumption habits of "conversation equals service" upstream, and this traffic interception directly bypasses Meituan's proud moat.

According to Sanyi Life's report, after Doubao connected with Douyin e - commerce and supported users to complete shopping orders within the app, it recently further completed the connection with Douyin's local life service. Users can use natural language in Doubao to express their needs, get recommendations for stores and group - buying packages, and directly complete the entire process of payment, order viewing, and verification code retrieval. Currently, categories such as food, movie tickets, and homestays are available.

This means that Doubao has built a complete "conversation equals transaction" closed - loop, which relies on the merchant resources and fulfillment system accumulated by Douyin's local life service over the years.

Meituan's ideal role is to be the "underlying service infrastructure" that all AI agents must call. However, the reality is cruel. ByteDance's Doubao and Alibaba's Qianwen are accelerating the construction of new "walled gardens" in the AI era, forming closed - loop business service systems and are not likely to easily open their call interfaces to external underlying services. The only AI assistant that is truly willing to complement rather than compete with Meituan is Tencent's Yuanbao.

QuestMobile data shows that Yuanbao had 57.35 million monthly active users in March and maintained a high penetration rate in first - tier and new first - tier cities. When users in Yuanbao casually say "find a non - spicy restaurant nearby", Xiaomei will be triggered to complete the recommendation and order placement. In this way, Meituan extends its service capabilities into users' daily conversations without waiting for users to actively think of opening the app.

This is a deal where Meituan exchanges "service capabilities" for "traffic entrances": using its scarcest assets (in - depth local life data and fulfillment network) to exchange for what it lacks the most (high - frequency AI conversation entrances).

The seemingly perfect complementary cooperation has three obstacles to overcome

Yuanbao's situation is more embarrassing than the outside world imagines. This embarrassment is not only reflected in the significant lag in data but also in the unclear positioning, and the value of being an independent entrance is constantly in question.

After the Spring Festival peak, Doubao's daily active users (DAU) remained at around 140 million, Qianwen's at 30 million, while Yuanbao's only at 9 million. Tencent invested more than 15 billion yuan in market expenses for Yuanbao in 2025, but not many of the acquired users stayed. This scale gap cannot be made up by the product iteration speed.

Since 2026, Yuanbao has taken many product actions, from Yuanbao's AI group chat to being the first to support Agent frameworks such as OpenClaw and Hermess. However, the only action that really made a splash in the market was the integration with Tencent's products such as WeChat, like the group chat summary function launched in May.

In terms of the ecosystem, Yuanbao has been embedded in Tencent's core products such as WeChat, Tencent Meeting, and QQ Browser in various forms such as plugins and contacts, becoming an "atomic" capability module. Different from the path of Doubao and Qianwen, which are striving to build independent AI entrances, Yuanbao's differentiated advantage lies in its deep binding with Tencent's user ecosystem.

However, this ecosystem binding also brings limitations: Yuanbao still needs to prove its value as an independent AI entrance, which cannot be directly given by the Tencent ecosystem itself. Doubao and Qianwen are becoming AI entrances that connect ByteDance's and Alibaba's e - commerce, group - buying, and other business ecosystems, while Tencent has almost no self - operated business and life service ecosystem, which is an inherent deficiency for Yuanbao in this regard.

This deficiency is particularly fatal in terms of commercial monetization. One of the biggest dilemmas currently faced by AI assistants is the high computing power cost and the difficulty in monetization. Even a wealthy company like ByteDance now has to consider launching a paid function for Doubao. For consumer - oriented AI assistants, simple chatting and tool functions can hardly cover the high computing power cost, and they must participate in real - world transaction scenarios to form a sustainable business closed - loop.

We can summarize the essence of this cooperation as follows: Meituan provides Yuanbao with the "body" (local life service capabilities), and Yuanbao provides Meituan with the "eyes" (front - end user entrances). However, it is not easy to make this seemingly perfect complementary cooperation work. It faces not only simple technical connection problems but also three core difficulties that are inevitable in cross - giant cooperation, and each one directly determines how far this cooperation can go.

First, where is the ceiling of the user experience? The technical form of this cooperation is "Agent to Agent". Yuanbao serves as the front - end conversation layer, and Xiaomei serves as the execution layer. The two complete task transfer through a standardized Agent protocol, and the users' original data will not be directly transferred between the two systems.

The advantage of this architecture is its clarity. Meituan doesn't need to open its core database to Tencent, and Tencent doesn't need to access Meituan's merchant evaluation system. However, the price of clear boundaries is limited capabilities. If a user says, "Help me find a restaurant suitable for a weekend dinner with children, with an average per - capita consumption of about 150 yuan, good reputation, and convenient parking", it is still unknown whether the standardized interface at the Agent protocol level can support answering such questions that require in - depth understanding of Meituan's structured data. Ultimately, the ceiling of the experience depends on how deeply the two systems are willing to connect.

Second, how to distribute the commercial interests. Each Meituan order completed through Yuanbao is essentially created by the traffic value of Yuanbao and the service value of Meituan. If the revenue - sharing ratio cannot be balanced, either party may lose the motivation for continuous investment.

Third, can the user experience be truly seamless? The integration of Qianwen and Alibaba's life services was completed within the same group, and the account system, payment link, and verification logic are naturally connected. The same is true for Doubao and Douyin's local life services. However, Xiaomei and Yuanbao are bridging two companies. When users issue instructions in Yuanbao, jump to Meituan to complete the payment, and encounter problems, who to contact for customer service? These connection details ultimately determine whether the experience is "usable" or truly "good to use".

This step is paving the way for WeChat's intelligent agent

When we place this cooperation in a broader context, its real significance becomes apparent.

On the day after Meituan released its financial report, the Financial Times of the UK disclosed a piece of news that ignited the market sentiment: Tencent is testing the prototype of an embedded AI intelligent agent in WeChat and has listed it as the company's top strategic priority. Tencent's stock price soared by more than 10% that day, the largest single - day increase since November 2022.

According to the Financial Times, the interaction entrance of this intelligent agent is very direct. Users only need to swipe right on the WeChat home screen to bring up the conversation window. After entering an instruction, the intelligent agent can automatically call on WeChat's massive mini - programs to complete various tasks, such as finding a coffee shop according to specific taste and price requirements and completing the order. Tencent plans to start the compliance approval this month. After completion, it will first conduct a gray - scale test for a small group of external users and then gradually promote it to all users in stages.

However, for this "super service hub" to truly stand up, a core question needs to be answered.

The core value of WeChat's intelligent agent lies in connecting millions of mini - programs. However, leading platforms such as Meituan, JD.com, Pinduoduo, and Ctrip in the mini - program ecosystem have their own data sovereignty and business interests and will not unconditionally hand over their service capabilities to WeChat's Agent for scheduling. Tencent cannot avoid a fundamental question: Why would big players be willing to connect?

The cooperation between Yuanbao and Xiaomei is precisely an early test of this question. Are leading Internet platforms willing to connect to Tencent's AI ecosystem in an "Agent to Agent" way? Meituan's answer is "yes". The interests of both sides are complementary. Meituan gets an entrance, and Yuanbao gets a scenario, and neither side is forced to give up its core rights and interests.

The logic of this "complementary cooperation" is the key reference for whether WeChat's intelligent agent ecosystem can recruit more ecosystem partners in the future.

In the future, the ecosystem pattern of WeChat's intelligent agent will probably be clearly divided into three levels.

The first level is "Agent to Agent". Platforms like Meituan, which have deep vertical data but lack independent AI entrances, connect to WeChat's Agent through the intelligent agent protocol, each retaining its data sovereignty and brand independence.

The second level is "Agent to mini - program". A large number of small and medium - sized merchants that do not have the ability to build their own Agents respond to the scheduling of WeChat's Agent by upgrading their mini - program interfaces. This is the level with the lowest cost and the widest coverage.

The third level is Tencent's self - operated businesses. Tencent's own products such as Tencent Maps, Tencent Health, and Tencent Video are directly and deeply integrated by WeChat's intelligent agent.

These three levels together form the complete picture of WeChat's intelligent agent ecosystem. One day before the Financial Times disclosed this news, the cooperation between Meituan and Yuanbao was announced during the earnings conference call. This overlap in time is not accidental. It seems that Tencent and Meituan are jointly sending a signal to the market: the cooperation at the first level has already begun.

Whether this sample can succeed will largely determine how far the boundaries of WeChat's intelligent agent ecosystem can extend.

The cooperation between Meituan and Yuanbao may not be large - scale, but it is crucial. It happened before WeChat's intelligent agent officially started and in the high - frequency transaction scenario of local life in China's mobile Internet. If this step works, there will be more stories to tell later.