AI Action Showdown: Alibaba Leverages Its Ecosystem Advantage

01

Will AI-Driven Shopping Disrupt Alibaba’s Current Business Model?

On the AI assistant battlefield, the focus of competition is shifting from dialogue-based interactions to the execution of instructions, and Alibaba has once again leveraged its ecosystem strength.

On January 15th, Alibaba’s QianWen App announced its full integration with key Alibaba ecosystem businesses, including Taobao, Alipay, Taobao Instant, and Fliggy. This enables users to perform AI-driven shopping functions, such as ordering food delivery, making purchases, and booking flights, directly within the QianWen App. This marks a significant step toward deep integration with the Alibaba ecosystem, following QianWen’s initial integration with Amap on December 18th.

Wu Jia, President of QianWen’s Consumer Business Group, emphasized at the launch event that AI is transitioning from conversational interaction to a new phase of autonomous task execution. He highlighted QianWen’s "unique advantage in combining the ‘Qwen’s strongest model’ with ‘Alibaba’s richest ecosystem.’"

QianWen’s upgraded capabilities focus on two key areas: the digital world and real-life scenarios. In the digital realm, the upgrade centers around the QianWen Task Assistant, which is currently undergoing targeted beta testing on both App and Web platforms. After testing concludes, it will be made fully and freely available to users.

However, the primary focus of this release is on real-life service scenarios. QianWen has announced its integration with Alibaba’s ecosystem capabilities, enabling an automatic planning and transaction execution closed loop that starts from consumer demand. This eliminates the need for users to break down their needs or switch between multiple apps.

At the launch event, Wu Jia demonstrated the AI food delivery function. He issued the instruction, "Help me order 40 cups of Boya Juexian from Overlord Tea Princess (Bawang Chaji)," and the QianWen App seamlessly called on Taobao Instant to place the order and completed AI payment directly within the app.

A natural question arises: Why is Alibaba integrating its ecosystem and enhancing AI’s task execution capabilities? Especially since AI-driven shopping’s direct purchase chain differs entirely from browsing on e-commerce websites, eliminating intermediate links and potentially disrupting the original advertising model of e-commerce platforms. Given this backdrop, why is Alibaba motivated to integrate QianWen with cash cow businesses like Taobao and Tmall?

According to Shuzhi Qianxian’s observations, on one hand, this is a crucial step for Alibaba to break through in the personal assistant space and establish a differentiated cognitive mindset. In the domestic personal assistant market, players like Doubao and KIMI had already launched a C-end talent acquisition war in early 2024. After Alibaba’s strong push of QianWen in late 2025, forming a differentiated mindset compared to earlier entrants and enhancing task execution capabilities is undoubtedly a key aspect.

Secondly, there are several paths to enhancing task execution capabilities. Objectively speaking, during the mobile internet era, strong barriers formed between domestic applications, and AI vendors face numerous obstacles in achieving cross-platform integration with other vendors. In contrast, integrating within a giant’s ecosystem faces relatively less resistance. Alibaba itself also possesses a rich ecosystem, providing ample space for such endeavors.

An industry insider believes that currently, QianWen’s user base is still niche compared to e-commerce user groups. For some time to come, there will be a long coexistence stage between traditional e-commerce forms and AI-allocated shopping traffic, and there will not be a strong impact on existing businesses in the short term.

At the same time, shopping demands in e-commerce scenarios are hierarchical, including both efficiency-oriented and browsing scenarios. AI-driven shopping’s efficiency primarily impacts the parts with clear goals and price comparison selections, while a significant portion of browsing demands will still be fulfilled based on the original forms. Meanwhile, if AI-driven shopping is not pursued by the giants themselves, entrepreneurial teams will step in. Alibaba’s self-integration of e-commerce processes is also an active transformation and self-innovation.

Shuzhi Qianxian tested scenarios such as cross-city travel planning and food delivery ordering and found that while the cross-app collaborative task execution capabilities brought by QianWen’s integration with the Alibaba ecosystem indeed provide convenience to users, the capabilities are still in their early stages of implementation. User habits, scenario demands, and capability boundaries are still being refined.

For example, in the cross-city travel planning scenario, traveling from a residential area in Hangzhou to a hotel in Shanghai for a meeting involves planning multiple routes and comparing time and travel costs. Traditional map software planning often defaults to Shanghai Hongqiao as the high-speed rail station, providing a usable but not optimal plan. With the large model assistant’s ability to call on map data, after pointing out errors, AI can quickly modify the plan and provide train schedule suggestions and reminders for the most ideal travel plan. This greatly resolves the cumbersome switching issues when operating across multiple apps.

The convenience of the AI food delivery ordering scenario is also evident. If users have clear preference demands, such as wanting to eat a clay pot stew, AI can recommend suitable nearby options from Taobao Instant in a card-based format, allowing users to switch between options by clicking the "next" label. This is more convenient than searching for stores one by one in a food delivery app and easier for decision-making.

However, using AI assistants for food delivery and completing other life services also requires long-term refinement based on user demands.

On one hand, current testing shows that Alibaba ecosystem integration is still in its early stages and requires further deep integration with QianWen. For example, when inputting a location in text, the model sometimes still has issues in accurately understanding the location.

Additionally, some vague purchase instructions cannot be well handled by AI yet. When viewing card-assisted decision-making in the large model dialogue box, it is not as convenient as on food delivery app or mini-program pages. Of course, the ability to handle vague instructions may improve with AI’s memory capabilities and deeper understanding of users, but this also indicates that there is still significant room for discussion in the product form evolution of large model task execution.

02

Enabling AI to Execute Tasks: Global Giants Choose Different Paths

Alibaba’s integration of its AI assistant with its mobile internet ecosystem within its system represents AI’s shift from dialogue to action. This attempt is not an isolated case, as global tech giants have made various attempts over the past year.

Overseas, Google has pursued two different paths.

Within the Google ecosystem, various core applications are integrating with Gemini, similar to QianWen’s integration with Alibaba’s ecosystem capabilities. Some believe that Gemini is no longer just an independent chatbot but has become the "AI hub" of the Google ecosystem.

For example, a few days ago, Gemini allowed secure connection to users’ personal Google app data, with Gmail and YouTube supporting connections. Meanwhile, at the Android system level, Gemini is gradually completely replacing Google Assistant. This path is also interpreted by the Zhiyuan Research Institute in the "Top 10 AI Technology Trends for 2026" as a characteristic of a super app adopting an All in One approach.

Google’s choice of this path, to some extent, is due to its late entry into the large model application space compared to OpenAI. Integrating with the ecosystem helps it quickly gain broader user touchpoints.

Outside the Google ecosystem, to call on external agent capabilities, a dedicated protocol is needed. In April 2025, Google launched the A2A open protocol specifically to address interoperability issues between AI agents. Industry observers note that when A2A was released, significant emphasis was placed on the B-end ecosystem, with all 50 ecosystem partners being B-end partners.

This means that Google’s two-pronged approach includes both internal ecosystem integration and external linking. Internal integration starts with C-end consumer applications, while the A2A open protocol focuses on addressing agent implementation within enterprises.

It is worth mentioning that two months after its release, the A2A protocol was donated by Google to the Linux Foundation, becoming a fully open-source, vendor-neutral community standard. The open-source move clearly transforms the A2A protocol into an industry infrastructure, facilitating intelligent agent connectivity and collaboration across the entire industry.

OpenAI’s exploration began with the formal launch of Operator in early 2025. As a startup, OpenAI previously lacked a strong ecological layout and had long been a market leader. Its exploration around task execution capabilities is more "AI-native."

Based on Operator, AI can view screenshots (visual understanding) + simulate mouse and keyboard operations to navigate web pages, highly similar to Anthropic’s "Computer Use" tool launched in late 2024. In July of last year, Operator was fully merged into ChatGPT’s "Agent Mode," supporting capabilities such as automatic form filling.

In September of last year, OpenAI collaborated with Shopify and Etsy, allowing U.S. users to directly purchase single items from Etsy sellers in ChatGPT dialogues. This is also an attempt at direct shopping within an AI dialogue box.

Since OpenAI lacks its own e-commerce ecosystem, accessing Shopify and Etsy capabilities requires completion based on agent protocols. External analysis suggests that this differs from QianWen’s integration with Alipay for AI payment. The latter, as an integration within the same ecosystem, enables smoother payment processes and the ability to perform complex tasks such as ordering milk tea and booking flights.

Doubao’s actions can also be analyzed from two levels. First is the Doubao App, which took a step forward in interconnecting with the Douyin ecosystem last year. In addition to connecting with Douyin’s content, it also connected with goods (commodities) in the Douyin Mall. AI became a shopping guide, helping users select and compare products. Of course, for purchases, Doubao did not choose direct in-app processing but still redirected to the Douyin App.

The "Doubao Phone," which debuted in early December last year and sparked tremendous controversy, differs from the above players’ attempts at the application level. It takes a "GUI" route at the mobile operating system level, being more direct and foundational.

Several days ago, Jin Hongbo, a senior industry researcher at the Zhiyuan Industry Research Center, interpreted during the release of the Zhiyuan Research Institute’s AI trend report that integrating AI assistants with a giant’s internal ecosystem can circumvent some barriers by integrating multi-industry APIs into a single application. To some extent, this represents a relatively steady reformist path and may be more practical.

The Doubao Phone, on the other hand, is an Agent OS centered around hardware, representing a radical attempt to break through application-layer restrictions. Although this path is the most disruptive, it faces the most challenges. For example, in addition to super app-layer restrictions, it also needs to penetrate and challenge the relatively entrenched pattern (landscape) of end-side hardware vendors. How to seize new user mindshare remains quite challenging.

However, some industry insiders believe that since the hardware-centered route may be more foundational, it could lead current giants integrating AI and internet ecosystems to reconsider manufacturing phones.

Regardless of how the situation evolves, it is foreseeable that competition around AI task execution capabilities will become a new focal point for giants in 2026.