Article | World Model Workshop
A week ago, Wang Xing said something during Meituan's Q1 2026 earnings call that many may not have fully grasped.
He said that in the future, Meituan’s service targets will no longer be limited to consumers (B2C) and merchants (B2B); serving AI agents (B2A) is becoming increasingly important.
To C and To B have been the fundamental business models of the internet over the past two decades.
But now, Wang Xing has introduced a new term: To A.
The truly radical aspect of this statement is that he didn’t treat the Agent as a tool, but instead redefined it as a customer.
If the agent is the customer, Meituan must consider how to make the agent more willing to promote Meituan, rather than just making users more willing to open the Meituan app.
Behind this, the distribution logic of the internet is being quietly replaced.
And the replacement speed is faster than imagined.
The Business Wars of the A Era
In the same week that Wang Xing introduced the "To A" theory, three other events occurred simultaneously.
Wang Xing announced on Meituan's earnings call that Xiao Mei AI Agent will collaborate closely with Tencent Yuanbao.
When a user says “I want to order takeout” in Yuanbao, Meituan takes over the subsequent steps of selecting a restaurant, placing the order, and delivery.
Just now, it was reported that JD.com has also partnered with Tencent on AI agents and has already engaged with terminal manufacturers such as Huawei, OPPO, and Honor.
Across the Pacific, OpenAI announced that ChatGPT will transform into a super app, integrating external services such as Booking, Spotify, and Expedia, with a focus on agents capable of autonomously completing various tasks for users.
Without any negotiation, it was as if a single signal had triggered a coordinated move—tech giants began forming alliances.
This is unusual.
Previous business battles: Douyin’s attempts to operate WeChat were directly blocked; Alibaba and Tencent mutually blocked links for nearly a decade; Meituan and Ele.me engaged in a brutal subsidy war that left both sides battered.
The culture of business competition on the internet has always been open and ruthless, with only one side surviving.
But this time, they are forming an alliance—and doing so quickly. Why?
The reason lies in Wang Xing's statement.
Today, when users open their phones, they open Meituan to order food delivery, Ctrip to book flights, and Taobao or JD.com to shop.
Each app maintains its own entry point, with traffic circulating within its own pool.
But once the Agent is integrated, users will only need to say one sentence: “Book me a Japanese dinner for tomorrow night.” The Agent will understand the intent, invoke the necessary services, and complete the order—all without opening any app.
This is where the big tech companies truly feel threatened.
If a user only speaks to the Agent in the future, are they still part of the chain?
Looking at the dense alliances between major companies this week, the logic becomes clear.
Meituan's Xiao Mei has integrated Tencent Yuanbao, while JD.com has partnered with Tencent to bring in Huawei, OPPO, and Honor—all essentially targeting To A and competing for agent recommendation slots.
So this isn't just a typical business rivalry—it's the arrival of the To A era, during which the internet is undergoing a new division of labor, and every company is trying to protect its own territory.
Different routes to A
It is worth noting that different companies have different approaches to To A. Currently, this round of major industry alliances is forming three distinct pathways:
Category 1: Super Entry Point + Service Provider.
Tencent Yuanbao, WeChat, and ChatGPT are all doing this.
Tencent Yuanbao integrates with Meituan, Tencent’s traffic channels connect to JD, and OpenAI integrates Booking, Spotify, Expedia, Canva, and others into ChatGPT—essentially, all of these are embedding services like food delivery, shopping, travel and accommodation, content, design, and payments into a single agent interface.
They are competing to be the first stop for users to submit their requests.
Category two: The app wraps itself as a service that can be invoked.
Meituan Xiao Mei, JD AI Agent, Taobao, Uber, Expedia, and OpenTable all belong to this category.
Their logic is practical: if users no longer open me directly in the future, I must at least ensure that I can still be invoked when an agent makes decisions on their behalf.
Better to retreat from the front-end interface to the back-end capability layer than to be completely bypassed.
Category three: Smartphone manufacturers serving as system-level agent entry points.
Huawei's Xiao Yi, Honor's YOYO, OPPO, and Xiaomi are taking a more fundamental approach.
They don't necessarily handle food delivery, shopping, or social media themselves, but they control access to the phone's system.
The user's first spoken words may first be picked up by the phone's AI assistant, then forwarded to WeChat, JD.com, and Meituan.
This is an opportunity for smartphone manufacturers to re-establish their foothold after losing their entry point in the app era, by leveraging agents.
However, some companies have taken a completely different path.
For example, Alibaba chose to first integrate internally.
Qwen, Taobao’s AI shopping assistant, Alipay’s AI wallet, and DingTalk’s AI travel services all integrate with Alibaba’s own services. Fliggy, Amap, and Taobao Flash Sales are also fully connected.
Alibaba chose to first transform itself into a complete closed loop, so that when it outputs externally, it does so as a bundled Alibaba service layer.
Regardless of the path chosen, everyone is competing to secure an irreplaceable position in the new To A chain.
Where is To A headed?
Alliance is the first step in this business war, but not the end.
The current situation appears to be a win-win for everyone:
Tencent Yuanbao has integrated the service capabilities of Meituan and JD.com, while Meituan and JD.com have gained access to WeChat’s traffic entry points. Smartphone manufacturers such as Huawei and OPPO have also joined in, benefiting all parties involved.
However, these partnerships have an inherent flaw: the interests of the entry party and the service provider are never fully aligned.
Can Yuanbao directly allow restaurants to register on its own platform tomorrow, bypassing Meituan?
Today, ChatGPT integrates with Booking; tomorrow, can it connect directly to hotel inventory without needing an OTA to mediate?
This is not a conspiracy theory, but rather a business temptation any platform with sufficient control over access points will face.
When WeChat launched mini programs, countless lightweight apps were absorbed into it; when Google Maps launched, local navigation apps disappeared en masse.
If a super app could truly enable suppliers to be directly orchestrated by their own agents—bypassing existing aggregation platforms for restaurants, hotels, ride-hailing, and shopping—then apps like Meituan and Ctrip would be in serious danger.
Although this path is difficult to achieve, it is not entirely impossible.
So today's alliance among major companies is essentially a race to secure a position before the window closes—better to be called upon than to be bypassed.
They bet that their service capabilities are strong enough to be irreplaceable, making the Agent entry point dependent on them.
In addition, this To A campaign has some other issues.
For example, could the Agent recommendation service eventually evolve into a new form of paid ranking?
If so, Meituan and JD would have to pay for the user traffic they originally owned, adding an extra layer of fees out of nowhere.
For example, if the Agent's recommendation results in an issue, who is responsible—the entry point or the service provider?
No one knows what the answer is, but everyone has already started running.
After all, in this To A restructuring, the most dangerous thing isn't falling behind—it's not hearing the gun go off.