Source | Bohu Finance (bohuFN)

Author | Qin Sanqian

This year’s Double 11 shopping festival has welcomed a new, special participant: AI.

AI can screen products, compare prices, stack coupons, and even proactively seek price protection refunds for consumers. For merchants, it can monitor inventory, analyze advertising effectiveness, and generate marketing materials in bulk. Complex and repetitive processes that once required constant back-and-forth between buyers and sellers are now being automated by AI.

On October 13, Alibaba launched its personal AI assistant, “Qianwen Go!,” which integrates shopping price protection, price comparison, and life service management into its AI capabilities. The following day, Alibaba released the office and e-commerce version of Qianwen, aiming to unify cross-platform business management for merchants through a single AI workstation.

Around the same time, Doubao Mobile Assistant faced a wave of collective pushback.

On October 14, Doubao Mobile Assistant announced the latest progress of its Screen Automation Operation Statement Protocol (SAEP), revealing that over 150 apps had explicitly refused to allow AI-driven automation. High-frequency apps like Taobao, Alipay, Meituan, and Pinduoduo were among those rejecting the protocol, while JD.com was notably absent from the list.

On one hand, e-commerce platforms are racing to integrate AI into the shopping experience. On the other, mainstream apps are determined to prevent external AI from interfering with their transaction processes.

AI is reshaping e-commerce, but each company is trying to secure a different foothold.

For Qianwen, e-commerce serves as the ideal entry point for personal AI assistants targeting consumers (C-end) and enterprise-level agents for businesses (B-end). For Doubao, shopping represents a core scenario for testing cross-app task execution capabilities. For JD.com and Pinduoduo, the value of AI revolves around traditional e-commerce metrics like product selection, conversion rates, and operational efficiency.

Behind these similar capabilities lie vastly different business strategies. This new competition centered around shopping carts is fully coming to light during Double 11.

01 Alibaba’s Dual Strategy

Alibaba’s AI strategy for Double 11 this year unfolds simultaneously on both the user and merchant fronts.

On the user side, Qianwen is evolving from a “shopping advisor” to a “shopping agent.”

In May, the Qianwen App fully integrated with Taobao, providing access to over 4 billion products on the platform. Users can now complete the entire shopping journey—from discovery and comparison to ordering and after-sales service—using natural language. The Taobao App also launched its own Qianwen AI shopping assistant, focusing on AI virtual try-ons and unified discount management.

Taobao and Tmall primarily operate as shelf-based marketplaces, where users refine their needs into keywords to search for and compare products. Now, Qianwen aims to redefine interaction logic. For example, when shopping for small-apartment furniture, users can simply describe their living space and budget, and AI will automatically match and filter options.

The “Qianwen Go!” test on October 13 took this a step further by intervening in shopping price protection, flight claim processing, and car maintenance inquiries. Shifting from “helping consumers find products” to “handling tasks for consumers,” Qianwen is transforming from a smart search bar into the primary gateway for expressing consumer needs.

This Double 11 logically serves as a testing ground for Taobao’s new “shopping agent,” with key performance metrics including order conversion rates, average order value, and repurchase rates. The complex rules of Double 11 make AI-driven optimization particularly valuable, as it can directly reduce calculation costs and recover orders abandoned due to complexity.

Moreover, the proven shopping festival represents a traffic peak for the Chinese internet. High-frequency e-commerce transactions provide an ideal environment for cultivating a super AI gateway, with user retention—the habit of entrusting daily tasks to AI—serving as the key metric.

Alibaba’s confidence lies in its closed-loop ecosystem. Taobao, Alipay, Gaode, and Fliggy form a complete fulfillment network. By helping Taobao close deals, Qianwen can establish itself as an all-powerful AI service gateway.

On the merchant side, the office and e-commerce version of Qianwen seeks to replicate this logic in the enterprise services sector.

Launched on October 14, Qianwen Office for E-commerce offers cross-platform business dashboards, trend insights, content generation, and cross-platform ROI optimization. It integrates with operational tools like Business Advisor, Qianniu, and Youzan. According to launch event data, e-commerce users already account for over 20% of its paid user base.

Multi-platform operations are now standard for merchants, but order, inventory, and advertising data are scattered across different backends. Qianwen Office aims to position itself as an independent third-party e-commerce agent, consolidating fragmented data into a single AI workstation.

The ambition behind this positioning is clear: if merchants grow accustomed to entrusting Qianwen Office with cross-platform reports, marketing plans, and inventory strategies, the accumulated operational context and workflows will create a highly defensible B-end software barrier.

While Qianwen Office’s core capabilities currently remain focused on data integration and decision support—with specific ad placements still limited by API permissions and manual confirmation—its strategic intent is unmistakable.

Alibaba aims not only to use AI to drive more sales on Taobao but also to embed Qianwen into the daily workflows of consumers and businesses through e-commerce scenarios.

02 Whose Wall Did Doubao Hit?

If Qianwen’s shopping capabilities are built within Alibaba’s proprietary ecosystem, Doubao Mobile Assistant is tackling a far more fundamental challenge:

Can users bypass apps entirely and directly assign tasks to AI?

In September, leveraging mobile system permissions, API interfaces, and GUI (Graphical User Interface) automation technology, the Nubia Doubao second-generation mobile assistant began interpreting user instructions to simulate manual clicks, swipes, and inputs, enabling cross-app execution of tasks like search, price comparison, and ordering.

This GUI automation model significantly lowers the barrier to utilizing existing apps but also places AI agents at the forefront of ecosystem conflict.

While simulated manual operations save users time, they easily trigger platform concerns over account permissions, payment security, and risk control. A deeper conflict lies in the redistribution of commercial traffic entry points.

Traditional e-commerce monetization relies on a “search—recommendation—browsing—advertising” pathway, where platforms generate revenue through exposure and traffic allocation. If consumers fully delegate their needs to external AI, product screening and comparison will occur outside the apps.

When external AI reorders products and redistributes traffic, e-commerce platforms risk losing their core advertising monetization mechanisms and recommendation algorithms. Platforms may still handle payments and fulfillment but will lose their most critical assets: user attention and traffic control. This is the root cause of the collective resistance against Doubao Mobile Assistant.

As mentioned earlier, as of October 14, over 150 apps had rejected Doubao’s SAEP protocol, including giants like Taobao, Alipay, Meituan, Pinduoduo, and WeChat.

It’s important to clarify that platforms resist external assistants’ unauthorized automated operations on their apps, not the technology itself. Taobao has its own AI assistant, and Pinduoduo offers in-app AI search. Each platform welcomes AI to enhance transaction efficiency on its own platform but firmly prevents external agents from eroding user relationships by bypassing rules.

For Doubao, shopping represents a litmus test for its cross-app agent orchestration capabilities. However, for transaction platforms, if this capability scales, apps risk becoming mere backend fulfillment tools. The more AI strives to become an independent super gateway, the greater the ecological resistance it faces.

This forces the industry to consider an alternative path: if AI doesn’t compete for independent entry points but instead empowers existing e-commerce transactions, would resistance be significantly reduced?

03 JD.com and Pinduoduo: A Different Transactional Approach

Among the apps resisting Doubao, JD.com presents a notable exception.

As of October 14, JD.com’s apps had not been explicitly added to the rejection list. Commercially, this suggests a possible strategic choice: if external AI drives new orders while transaction fulfillment remains under the platform’s control, moderate openness is acceptable.

Xiaohongshu, once mocked by the industry for its “open-loop” strategy, provides a relevant precedent. During its efforts to cultivate an e-commerce ecosystem, Xiaohongshu repeatedly opened and closed external links. Its “Grassroots Direct” feature, for example, allowed advertising notes to link directly to external e-commerce platforms like Taobao and JD.com. This approach—sacrificing some front-end decision-making traffic for cross-platform orders and conversion gains—represents a compromise between traditional traffic platforms and fulfillment platforms.

JD.com’s long-established self-operated supply chain and logistics fulfillment system form its strongest defense. Even if users compare prices via external AI, final purchasing decisions still hinge on price, inventory, and delivery speed. External AI stimulates and filters demand, while JD.com handles fulfillment—a logic that remains intact.

Of course, JD.com is also building its own AI defenses. On the user side, it upgraded its AI shopping assistant “Dongdong” and launched JoyAI real-time video digital humans, integrating shopping with lifestyle services. For merchants, it introduced the Jingmai AI Operations Center, planning to provide digital human livestreaming services to over 90,000 merchants during Double 11.

JD.com’s AI strategy is straightforward: focus on services and fully integrate into existing supply chain and transaction processes to reduce friction for both buyers and sellers.

Compared to JD.com’s high-profile displays, Pinduoduo’s approach is more pragmatic and subtle.

In May, Pinduoduo quietly launched its AI search function. When users input complex long-tail requests like “find low prices, specific brands, black-label products, or billion-subsidy items,” AI precisely dissects the conditions and matches them with its product library. Prior to this, the platform had already rolled out C-end features like virtual fitting rooms, AI friend interactions, and AI interactive dramas to boost engagement and aid decision-making.

Pinduoduo positions AI purely as a highly efficient internal tool. Whether for product distribution, engagement, or retention, each new feature serves a clear purpose without disrupting users’ accustomed shopping habits. This contrasts sharply with industry attempts to reinvent shopping habits through conversational “shopping guides.”

Rumors suggest Pinduoduo still “adheres” to traditional programming methods internally. The company cares less about whether employees use tools like Vibe Coding to boost efficiency and more about actual output. This extreme efficiency-driven approach mirrors the strategy employed by Zhao Jiazhen, Pinduoduo’s current co-chairman, when he initially built the agricultural product supply chain.

On the merchant side, Pinduoduo’s official open platform has long provided merchants with access to order and logistics interfaces, allowing third-party service providers to develop AI customer service and product management tools under authorization. One Pinduoduo merchant revealed that platform representatives even encourage the use of ByteDance’s e-commerce content generation tools and provide hands-on guidance.

The pragmatic paths taken by JD.com and Pinduoduo demonstrate that AI doesn’t need to disrupt consumer interaction gateways immediately. Instead, it can start by addressing specific, time-consuming transaction processes.

E-commerce is an extremely results-driven industry. If AI generates 10,000 marketing images for merchants but fails to improve conversion rates, its commercial value diminishes. Similarly, if AI compares prices ten times for consumers but delivers a poor final experience, repeat purchases are unlikely.

Helping merchants manage their businesses and making decisions for users—this Double 11, the competition among e-commerce AI has shifted to the “agency layer.”

Yet regardless of how front-end entry points evolve, product supply capabilities, fulfillment speed, and after-sales support remain the cornerstones of successful transactions. AI can alter traffic flow but cannot overturn fundamental supply-and-demand laws.

The best AI is the one that facilitates transactions most efficiently.

The cover image and illustrations for this article are owned by their respective copyright holders. If copyright owners believe their work is unsuitable for public viewing or should not be used without compensation, please contact us promptly, and our platform will make immediate corrections.