On the evening of August 10, the Qianwen App unveiled two simultaneous updates.

One update introduced paid membership tiers for its office assistant, offering three options: Premium, Elite, and Flagship. Monthly fees range from 19 to 128 yuan, with an annual maximum of 1,499 yuan. Video generation quotas are billed separately, with a limited-time offer of 10 videos for 26 yuan (regular price: 45 yuan for 10). The other update was the launch of an open platform, integrating AI agents from over a dozen external companies, including SF Express, Ziroom, FlashEx, and Yingmi Fund. Now, users can send parcels, rent cars, or book housekeeping services within Qianwen by simply using the @ symbol.

The first update charges users, while the second opens up to third parties. These two moves reflect a shift in Alibaba's approach to C-end products.

Over the past two decades, Alibaba's core strategy for C-end products has centered on facilitating transactions, integrating users into the Taobao ecosystem with a fully closed-loop system that encompasses search, ordering, payment, and logistics. Revenue is generated through transaction commissions and advertising fees.

Qianwen, at least for now, is attempting to break away from this model. Services invoked by users within Qianwen are no longer confined to Alibaba's ecosystem. SF Express's delivery services, Ziroom's rental listings, and Yingmi Fund's financial management tools mean that transaction data and user relationships do not fully belong to Alibaba. Alibaba's approach to C-end products is shifting from incorporating more users into its own ecosystem to connecting users with more external services.

The recent updates on paid services and open platforms represent a focused implementation of this shift.

I. Qianwen App Adopts a Dual-Track Strategy

The two updates Alibaba has simultaneously implemented on Qianwen target different business objectives, each with its own strengths and limitations.

The path of office-related fees offers a short-term monetization route with high certainty.

Qianwen's office assistant leverages the latest Qwen Flagship model, supporting remote mobile operations and directly delivering files such as web pages, PPTs, Word documents, and Excel spreadsheets. By tapping into the essential office scenario, there is a clear willingness among users to pay.

Wu Yongming has publicly stated that B-end clients have a stronger payment willingness and easier ROI calculation, while C-end “requires a certain investment cycle.” From this perspective, the office scenario is the C-end sector closest to the B-end payment logic. Video generation, purchased on-demand, covers elastic demands for creation and entertainment. Separate billing helps clarify payment willingness and pricing flexibility across different scenarios. Unlike Doubao, which launched paid services earlier and bundles office and video features, Alibaba's choice to charge separately may reflect a more cautious approach to refining user personas on the Qianwen App.

However, the ceiling for this path is clear. The office narrative is not particularly exciting, with limited market space. For a giant like Alibaba, focusing solely on the office scenario seems insufficient for growth. While office fees can cover some costs, they are unlikely to support the entire commercial goals and capital narrative of the Alibaba Group.

Previously, Wu Yongming set a target for the cloud and AI businesses to achieve commercial revenue exceeding $100 billion within five years. Achieving this scale would be difficult to cover solely through office subscriptions.

The open platform path, on the other hand, bets on the other end, embodying Alibaba's idealistic vision for AI narratives.

After third-party agents are integrated into Qianwen, users can complete tasks such as sending parcels, renting apartments, or consulting financial services without leaving the app. Qianwen provides infrastructure like accounts, AI-powered payments, and order access, with standardized protocols significantly lowering the migration threshold for external agents.

The advantage of this path is its strong AI-native nature, aligning with external expectations for big tech's AI-to-C products—more scenarios attract more users, more users generate richer data, richer data feeds model iteration, and stronger models attract even more users, creating a virtuous cycle.

The problems are equally clear. The monetization path for super entrances in the AI era has yet to be proven. Integrating third-party agents into Qianwen is not technically difficult; the real challenge lies in attracting and retaining users from these platforms onto the Qianwen App, ultimately achieving commercial conversion.

Currently, for many third-party apps, Qianwen may serve as an additional traffic channel but is far from being their primary platform. This may be one reason why third-party apps are willing to open their interfaces to Qianwen.

Despite the pros and cons of both paths, adopting a dual-track approach is almost an inevitable choice for Qianwen. This dilemma is not unique to Qianwen but is shared by all entrance-type products.

Focusing solely on the office scenario cannot bridge the gap, necessitating the attraction of more users through a super entrance. Conversely, betting solely on the entrance cannot cover current investments, requiring office fees to provide cash flow support. A short-term inability to choose between the two leads to parallel exploration: one path tests the limits of the super entrance, while the other explores a quick monetization route to offset significant infrastructure and talent investments. The two are not mutually exclusive.

What Alibaba has done relatively well is translating its long-discussed open ecosystem into reality. Not only are Alibaba's own Cainiao and Kuake Cloud Disk integrated, but services from non-Alibaba ecosystems like SF Express, Ziroom, and FlashEx are also included. This goes beyond mere conceptual declarations.

From a broader perspective, while the two updates introduced by Qianwen are only phased actions, they clearly reflect a shift in Alibaba's mindset regarding C-end products.

II. Qianwen Moves Beyond Its 'Alibaba Identity'

In the internet era, Alibaba's core objective for C-end products has been to facilitate transaction closures. Launched in 2003, Taobao ensured users remained within the Alibaba ecosystem throughout the entire consumption journey, from search and ordering to payment, logistics, and confirmation of receipt.

Throughout the consumption chain, Taobao acted as a shelf, Alipay addressed credit issues, and Cainiao improved efficiency.

The core idea was to incorporate core assets such as users, data, and transaction chains into boundaries defined by Alibaba, completing consumption on its own shelves. After merchants joined Taobao, they relied on Alibaba's traffic, payments, and rules, with Alibaba earning transaction commissions and advertising fees accordingly. Users, however, remained relatively passive participants in this chain.

This system operated for two decades, enabling Taobao to achieve an annual GMV exceeding one trillion yuan.

What Qianwen is doing now fundamentally differs from this approach, even dismantling the barriers previously erected by Taobao and Tmall.

When users send parcels via SF Express, rent apartments through Ziroom, or manage finances with Yingmi Fund within Qianwen, these services are not directly provided by Alibaba. Transaction data and user relationships do not directly accumulate on Alibaba's platform.

The Qianwen App functions more like an agent service hub. When users raise demands, it understands their intentions, assigns tasks to corresponding agents, and provides infrastructure like accounts, payments, and orders. Services can be directly provided by SF Express or Ziroom, with Qianwen's value realized only during the demand fulfillment process, measured by Token usage.

The intent behind Qianwen's recent updates becomes clear. Under the unified goal of pursuing higher Token usage, the office assistant can quickly gain user stickiness and commercial conversion, while the open ecosystem allows Qianwen to extend more tentacles outward, connecting more agents and scenarios to ultimately reach more users.

To some extent, Alibaba is shedding its previous dominant posture, shifting toward a more user-centric product model. Whatever users need, Qianwen will coordinate, regardless of whether the service comes from within or outside Alibaba. This posture might have been hard to imagine from Alibaba two years ago.

However, for Alibaba to truly complete this shift, it may need to overcome a series of internal and external resistances.

Whether the office assistant can successfully generate revenue remains subject to organizational-level negotiations. On August 3, Qianwen Office, an independent product formed by integrating QoderWork, Wukong, and MuleRun, began public testing under the oversight of DingTalk CEO Chen Yusen.

The Qianwen App falls under the jurisdiction of Group Vice President Wu Jia. The introduction of a paid membership for the office assistant undoubtedly encroaches on the core territory of Qianwen Office. Some media outlets have sharply pointed out that the coexistence of the two lines is “only temporary.”

A deeper issue with the open ecosystem lies in a series of data barriers. The depth of the data pool determines the speed and effectiveness of model iteration, which fundamentally determines whether the Qianwen App can hold its ground amid competition from Doubao and WeChat.

In the mobile internet era, Alibaba relied on its self-built transaction system to construct a powerful user database, with users' search, purchase, and return records accumulating within Alibaba's data system, reinforcing recommendation capabilities. Now, as a service hub, external agents' data flows through Qianwen, but it is difficult to deliver all data to a single coordination platform.

Previously, the full integration of Taobao and Qianwen was driven by strategic-level initiatives from Alibaba's top leadership and still involved repeated negotiations. If third-party apps are to open core data interfaces for Qianwen as a service hub, technical resistance is likely far less significant than interest-based resistance.

When Qianwen integrated with Taobao, Wu Jia stated, “Whether ordering within Qianwen, Taobao, or Taobao Flash Sale apps, it all happens within the Alibaba ecosystem, which is the same for us.” However, for SF Express, whether users place orders through SF Express or Qianwen may not be the same.

The battle for AI-to-C is converging on the same battlefield, with each player starting from different points. Office agents compete for commercial gains, while open ecosystems compete for user attention. Both have become essential battlegrounds for big tech. Going forward, the number of integrated agents will no longer be a barrier; the key differentiator will be who can truly retain users behind these services.

During this year's Spring Festival, Qianwen invested 3 billion yuan to treat the nation to milk tea, temporarily boosting DAU to 73 million. However, the issue of user retention after the red envelope surge remains a constant Sword of Damocles hanging over the app. According to QuestMobile data, the average 30-day retention rate for mainstream AI apps in China is just 12.8%. Until the issue of user retention is resolved, the paid foundation of these agent products will remain shaky.

With a $100 billion revenue target looming, Alibaba cannot afford to wait until everything is perfectly prepared before moving forward.

Qianwen's true enemy may not be Doubao or WeChat but Alibaba itself. The organizational inertia shaped by 20 years of closed-loop thinking is being gradually dismantled by Qianwen. Whether Alibaba can rewrite its history of C-end product development with Qianwen may depend on its willingness to accept deeper-level changes.

In the agent era, few possess an absolute first-mover advantage. Users hold more choice, and if a product falls short, they will quickly vote with their feet. This means that, in the short term, companies must “please” users to some extent—a challenge perhaps more difficult than any technical hurdle for a company accustomed to controlling everything.