If 2023 is the year of global AI, 2026 should be the year of implementation of AI applications in China.
In May 2024, Tsai Chongxin and Wu Yongming issued the first letter to shareholders after Alibaba’s new management team took office. “No industry will be immune to the disruptive impact of AI over the next 10 years,” the letter said.
Two years have passed, and Alibaba has gradually become one of the platforms with the highest investment and deepest layout in China’s AI industry.
According to the “China AI Cloud Market Share 2025” report released by Omdia, Alibaba Cloud ranks first in both AI IaaS and MaaS (MPS core submarket), further strengthening its leading position in China’s AI infrastructure field.
On May 20th, Alibaba Cloud announced at the summit that it had completed the upgrade of “Chip-Cloud-Model-Inference” to become a full-stack agent, and at the same time launched a new AI product official website “Qianwen Cloud” equipped with a super node server equipped with the self-developed AI chip Zhenwu M890 and the latest flagship model Qwen3.7-Max.
From underlying chips and cloud computing to large-scale models and agent platforms, Alibaba is rapidly enriching full-stack capabilities for the AI era.
Meanwhile, Yilan business realized that: Alibaba has been trying to integrate all of its operations with AI over the past year, but the launch of AI products for some companies seems a little hasty.
Alibaba’s e-commerce: Completely online powered by AI
According to incomplete statistics from Yilan Business, almost all of Alibaba’s core e-commerce and lifestyle services businesses intensively launched AI products from 2024 to 2026.
Based on the actual implementation of AI in multiple companies such as Taogongchang, 1688, and Tmall Supermarket, Alibaba’s AI exhibits the following characteristics:
First, the depth of AI adoption has essentially evolved into “intelligent agent hosting.”
Early AI products, such as the intelligent publishing function and intelligent hosting function that Xianyu launched in 2024, were like auxiliary tools to simplify operational processes. However, from the end of 2025 to 2026, Taogongchang’s “Spark 3.0” built an architecture of “1 AI store manager + N agent assistants” and realized a full hosting model. Tmall Supermarket’s “Supercat AI Supermarket Agent” covers 16 business sub-agents, including business diagnosis, advertising management, and intelligent replenishment.
Second, the scope of AI has also expanded from a single “product distribution” to a closed loop of decision-making and services.
In addition to releasing the AI version of the app, 1688 also independently released an enterprise search and research tool, “88 Search,” and an “AI Digital Employee” merchant assistant. Fliggy’s “Ask” integrates multiple professions such as itinerary assistants, route planners, and hotel consultants to form a multi-agent-driven travel service. Tmall’s ‘Designer’ enters directly into the home improvement market, offering a full set of capabilities from AI design to supply chain and construction fulfillment. AI is no longer just a button, it’s the operating system underlying your entire business.
Finally, the introduction of AI in e-commerce business primarily focuses on improving the “transaction efficiency” of merchants.
Taogongchang’s Spark, 1688’s digital employees, Tmall Supermarket’s agents, or Ele.me’s AI payment intelligent manager…they all tackle issues related to “goods distribution” such as product selection, pricing, product launch, advertising, customer service, replenishment, supply chain management, and more. This reflects Wu Yongming’s statement that “C-end AI is the creator of token consumption scenarios.” Only by enabling merchants to first use AI to achieve efficiencies can we drive a larger data flywheel and upgrade the experience on the user side.
However, most of these AI products are currently only in the testing phase. A relatively mature platform is one that is closer to the industrial end, such as the 1688. Products like Xianyu’s AI photography and Tmall Supermarket’s intelligent agent have only been on the market for a few months, so they will need time to establish user habits and validate their business models.
Further testbeds: reworking niche scenarios
In addition to e-commerce, Alibaba has thousands of scenarios suitable for reworking with AI. Yilan business counted the most typical ones.
DingTalk is Alibaba’s carrier for the enterprise (Level 1) layer in the AI era. Its product line iterates at a pace of major versions every six months. First the agent entry point (AI assistant) is created, then the underlying system (agent OS) is built, and finally it is transformed into an enterprise-level agent platform. DingTalk has hundreds of millions of enterprise users and tens of millions of organizations, which naturally has high-frequency, high-value data scenarios. These are exactly the “fuel” needed for AI training and inference. Outside of e-commerce, DingTalk can be said to have the most potential as a “data flywheel” among Alibaba’s AI businesses.
There is also Gaode Map, which has been very popular recently. Gaode’s unique value lies in the fact that its location and movement data are frequently the only point of entry into the “offline physical world” of the Alibaba ecosystem. Whether it’s food delivery, taxi hailing, local living, or sightseeing, Gaode is the starting point for your decisions. Therefore, Gaode’s AI upgrade first focused on intelligent agent features such as AI navigation and AI instant service. By May 2026, this was further upgraded to the ‘Gaode Spatial Intelligence Open Platform’, connecting a closed loop of spatial awareness, intelligent decision-making and scenario-based services.
In contrast, within the Alibaba ecosystem, Huijing Entertainment’s AI scenarios are more focused, but with relatively limited room for imagination. Recently, there were reports that the Miaoya Camera team under Huijing Entertainment has disbanded. Huijing Entertainment’s digital human projects represent the company’s AI transformation, but Yilan Business observed that most of these projects are being operated on a project-by-project basis, with some digital humans on hold.
Currently, the Damai client is more recognizable to users, but it only has basic features such as AI intelligent recommender and multi-intent recognition.
In addition to these internal businesses, Alibaba is making three more moves in its C-end AI entrance layout. They are Quark, Qianwen, and UC Browser, which follow completely different paths.
Among them, Qianwen is Alibaba’s biggest bet on independent AI-native applications. According to QuestMobile data, as of March 2026, Qianwen’s monthly active users reached 166 million, ranking it second after Doubao among similar products. Unlike most AI assistants, which are still focused on Q&A, writing, and searching, Qianwen’s difference lies in “getting things done.”
In January 2026, Qianwen App announced integration with Alibaba ecosystem businesses such as Taobao, Alipay, Taobao Flash Sale, Fliggy, and Gaode, seeking to repackage the Alibaba ecosystem’s transaction, payment, fulfillment, and service capabilities into an AI gateway.
Quarks play another role. AI transformation is progressing based on high-frequency tool scenarios such as search, learning, office work, and file processing. Compared to Qianwen, which is a strategic bet on the “gateway of the future,” Quark is more of a mature example of Alibaba accelerating AI penetration in its existing traffic pool. The value lies in testing whether AI can be incorporated into the searches and tool links that users are familiar with to increase usage and retention.
UC Browser has chosen a more lightweight path. UC Browser launched the “AI group chat” function in January 2026. Default members include UC’s native AI assistant “Xiaoyou,” Quark AI, Tongyi Qianwen, and external collaboration DeepSeek, and employs a multi-agent architecture. Users can ask multiple agents the same question. This feature currently has a deep entry point, is not widely promoted, and is more of a small-scale experiment by Alibaba with traditional traffic entry points for browsers.
Overall, compared to the e-commerce side, Alibaba’s non-e-commerce side AI applications cover a wide range of areas, from corporate collaboration to retail operations, travel navigation to culture and entertainment consumption, touching on almost every aspect of lifestyle services.
Alibaba’s four tests of AI: Efficiency, Users, Ecosystem, and Entrance
In fact, after three years of development, China’s AI industry has completely transitioned from a simple long-distance competition in model technology to a fierce “positioning competition” centered on application implementation.
The hallmarks of this competition are a high concentration of capital, a rapid expansion in the scale of users, and fierce competition between giant companies. According to the data, in 2025 alone, the amount of financing for domestic AI application fields will exceed 100 billion yuan, and the number of monthly active users will exceed 446 million. During the 2026 Spring Festival, major companies such as Tencent, Alibaba, and ByteDance spent more than 8 billion yuan on hongbao to gain entry. Ace products such as Doubao, Qianwen, and DeepSeek all have more than 100 million monthly active users, forming differentiated competitive paths such as “traffic × experience”, “ecosystem × closed loop”, and “technology × word of mouth”.
Therefore, from a higher perspective, the core purpose of Alibaba integrating all its operations with AI is to gain a foothold in this fierce “positioning competition” centered around application implementation. At the same time, for Alibaba, rushing to integrate all of its business with AI is also a way to force companies to undergo a degree of four-way stress testing.
The first test is whether AI can deliver real business outcomes. If AI cannot be translated into quantitative business revenue, it is just an expensive technological gimmick.
The second test is user acceptance. The value of AI must be realized through continuous use by users, and the premise of user usage is that the product is truly integrated into their daily decision-making processes. It’s certainly exciting to be able to create an AI-native application with over 100 million monthly active users, but the key to commercial success is solving the problem of getting users willing to pay.
The third test is ecosystem synergy. Alibaba has a variety of scenarios, including e-commerce, payments, logistics, local life, office work, and travel. However, history has shown that consolidating internal traffic pools is never easy. Whether AI truly breaks down organizational walls and enables the seamless flow of data and services within ecosystems will directly impact whether we can realize the vision of “one AI gateway to manage everything.”
The fourth and most fundamental test is: Where will the entry point be for AI? Alibaba’s bets on multiple avenues at once accurately demonstrate that the entire industry is still searching for answers.
The full-stack layout gives Alibaba confidence to stay in the game, but the real deciding factor lies in who can first establish the positive flywheel of “Improved AI efficiency – Merchant profits – User retention – Continuous token consumption.” There are no standard answers to this test, and scores are determined by time.
This article is from the WeChat public account “Yilan Business” (ID:yilanshangye), written by Li Yan, and published by 36Kr with permission.
