Alibaba's AI assistant app, "Qianwen," announced a comprehensive update integrating with Alibaba's Taobao, Alipay, and other Alibaba ecosystem services, including Ele.me (via Taobao Flash Sale), Fliggy, and Amap. This means users can now directly use natural language within Qianwen's chat window to complete tasks such as ordering coffee, booking flights, and checking routes, officially transitioning from "information retrieval" to AI-assisted "transaction execution."
This wave of AI application upgrades signifies that agentic AI is moving from a technological concept to commercial application. Not only Alibaba, but also giants like ByteDance, Google, and OpenAI are vying for this "super gateway" to AI that can truly generate commercial value.
Qianwen's Strategy: The Advantages of Alibaba's Ecosystem Service Integration
The core logic behind Qianwen's latest update is simple: "Connectivity".
According to actual testing, when a user tells Qianwen "Order me a coffee," after authorization, Qianwen automatically captures the user's location, recommends nearby brands (such as Luckin Coffee), and directly generates a payment link, redirecting the user to Alipay to complete the payment. If the user says "I want to book a flight to Beijing tomorrow," Qianwen can also directly call Fliggy's API to list flights and provide a booking card.
The advantage of this model lies in Alibaba's vast service ecosystem (Taobao, Gaode Maps, Fliggy, etc.). Through deep integration of internal APIs, it can ensure order success rates and data immediacy. This is a "front store, back factory" approach, using service certainty to enhance the practical value of AI.
ByteDance x ZTE: The "Brute Force Hacking" of the Doubao Phone
Compared to Alibaba's ecosystem integration, ByteDance's "Doubao" has taken a more geeky approach— Auto-UI (Automated Interface Operation).
ByteDance's recently launched Doubao phone, developed in collaboration with ZTE, features an AI assistant that doesn't rely on open APIs in apps. Instead, it uses a visual model to recognize pixels and buttons on the phone screen, simulating human finger taps and swipes. If you say "Help me hail a taxi," Doubao will directly "understand" the Didi app interface and complete the subsequent operations.
The advantage of this approach is that it is not limited by whether the app opens its API, and theoretically it can take over all mobile applications. However, the challenge lies in accuracy and cross-application stability, and it requires breaking through the permission restrictions of the underlying mobile operating system. It is a battle for a "super operating system" with a bigger stakes.
However, Doubao Mobile's approach has also begun to face challenges. Many applications and games have started to restrict Doubao Assistant AI from "intruding," and some have even directly blocked user accounts, believing that Doubao Mobile's assistant service has completely undermined the security mechanism.
AI Agent Deployment by Global Giants
Globally, the war of "AI getting things done" is raging.
• Google Gemini x Walmart:Google has partnered with retail giant Walmart. Based on the UCP (Universal Commercial Protocol), Gemini can directly access Walmart's real-time inventory, allowing users to complete shopping and payment directly within a conversation, and even arrange pickup. This is strikingly similar to Alibaba's model, both being powerful alliances.
• OpenAI Operator:OpenAI launched last yearA browser proxy product called "Operator"The goal is to enable AI to automatically navigate web pages and complete complex tasks such as booking tickets and shopping. This shows that the simple SaaS subscription model has reached its ceiling, and entering the transaction flow is the key to monetization.
• Microsoft and Amazon:Microsoft is integrating the Office ecosystem through Copilot, focusing on automating office scenarios, while Amazon continues to optimize the integration of its Alexa with the shopping process.
Analysis: The next steps for AI are "trust" and "privacy"
From "chatting companions" to "getting things done," the commercial value of AI has increased significantly, but risks have also arisen.
While Alibaba's Qianwen model offers a smooth user experience, it involves payment and personal data authorization. Ensuring the AI doesn't "buy the wrong thing" or "leak privacy" is a major concern for users. Doubao's Auto-UI mode, which directly takes over the phone's screen, will face even stricter scrutiny regarding its permission management and security.
Analysts point out that for AI to achieve widespread adoption and monetization, in addition to technical accuracy, it also requires the integration of the upstream and downstream of the industry chain and the distribution of benefits, which will be a long process. Currently, Qianwen and Doubao are still in the stage of "integrating internal resources." The real final battle will be who can build a universal AI agent that is cross-platform, cross-ecosystem, and that users truly "dare to use."
For consumers, there may not be a rush to hand over their credit cards to AI in 2026, but trying to use it to compare prices and check routes can certainly save a lot of time switching between apps. The "sweet spot" of AI applications is slowly emerging.
