Tencent WorkBuddy, can't learn to work the way Doubao does?

Will Tencent's WorkBuddy integrate WeCom?

On the day ByteDance's Doubao Work was launched, Tang Daosheng, CEO of Tencent Cloud and Smart Industries Group, released a lengthy internal letter.

The Chatbot battle is already a thing of the past, and AI office agents are the current hot spot. Tang Daosheng mentioned the competition between Yuanbao and Doubao over the past year, acknowledging that the rival is worth learning from. Under huge pressure of user growth, Yuanbao spent plenty of energy on promotion and traffic diversion, while the model and product were not ready at that time, so the actual performance failed to meet expectations.

What Tang Daosheng mentioned more is WorkBuddy, Tencent's most popular AI office product at present.

Especially when talking about organizational structure, he noted that the team is pushing for interconnection between WorkBuddy and products including WeCom, Tencent Docs, ima, Tencent Meeting and iwiki, so that high-frequency demands such as document processing, knowledge management, meeting summarization, collaborative work and human-AI co-creation can be realized in one stop on WorkBuddy.

The purpose of this move is to focus resources on in-depth scenarios, to "continuously acquire more real task data, user interaction feedback and scenario-based context, and apply these feedback to tuning, capability complementation and experience optimization."

The centralized organizational layout that enables one-stop access to multiple high-frequency office demands is exactly the key difference between Doubao Work and WorkBuddy at the current stage.

Doubao has merged Feishu, allowing the Agent to truly integrate into internal organizational workflows. It later absorbed two teams from Coze and Trae Work, becoming the product with the highest concentration in the AI office sector, under the unified management of Zhao Qi. Meanwhile, Alibaba's Qwen Office has also achieved ToB integration, incorporating three Agent products QoderWork, Wukong and MuleRun, which are uniformly overseen by Chen Yusen.

In comparison, WeCom as an IM product belongs to Tencent's Weixin Group, while products including WorkBuddy and Tencent Docs fall under Tencent Cloud and Smart Industries Group. The organizational barrier between IM and Agent products has brought challenges and tests to the ecological collaboration of Tencent's AI office products.

Therefore, in the face of a series of integration and catch-up moves from ByteDance and Alibaba in the AI office field, can Tencent's WorkBuddy, which has millions of daily active users, maintain its first-mover advantage? Both Feishu and DingTalk have transformed, can WeCom stay unchanged? If competition further intensifies, can WeCom and WorkBuddy achieve deeper collaboration or even full integration?

After all, the office Agent track has already become a red ocean. In the past, these products were only intelligent tools, but now real organizational relationships and contextual data form the competitive moat.

The login box defines the worldview

In its Q1 financial report this year, Tencent described WorkBuddy as "the most widely used productivity AI agent service in China", with its first-mover advantage very prominent, boasting millions of daily active users. Entering August, Alibaba's Qwen Office and Doubao Work were released one after another.

The AI office trend has been verified, as Agents can independently execute more complex work tasks. Under this premise, how will ByteDance and Alibaba compete in the office market?

Doubao Work has two main login methods: Feishu account and Doubao account. Qwen Office also has two main login methods: mobile phone number and DingTalk account. In contrast, Tencent's WorkBuddy has only one most prominent login method: WeChat.

WeChat login reflects WorkBuddy's most natural inertia on the C-end, but the greater inherent advantage of Doubao Work comes from Feishu rather than Doubao itself.

Logging in to Doubao Work with a Feishu account means it will inherit all enterprise assets within your permission scope, including group chats, meeting minutes, documents, schedules and more, enabling it to deeply understand users' work context. This is the biggest difference between Doubao Work and Tencent and Alibaba's AI office products. Meanwhile, the original Doubao Enterprise Edition entrance inside Feishu has also been upgraded to Doubao Work.

In other words, Doubao Work aims to make Agent the new entry point for work, and the current trend is that office software is no longer the only entry point for production scenarios.

In contrast, WorkBuddy is not a built-in function of any office suite, but an Agent that runs independently on users' computers. In terms of permissions, it cannot access user chat records and context inside WeCom unless users manually provide relevant files. Essentially, they are two separate products that are only connected to WeCom.

The "Channel Access" section in the WorkBuddy Enterprise Edition document clearly states that it "supports integration with WeCom AIBot, QQ Robot, Feishu and DingTalk to receive and reply to Agent messages". Users can manage the instant messaging channels of Agent in the enterprise backend, but for WorkBuddy, WeCom, Feishu and DingTalk are in a parallel relationship, with no deep binding to any single platform.

This also means WorkBuddy wins with a wider coverage, while Doubao Work, to carve out its place in the AI office track, will definitely focus on exploring scenario depth.

During the live broadcast on the day Doubao Work was released, the AI capabilities demonstrated with focus were all about task processing empowered by Feishu collaboration, including real-time content quality inspection, remote content production and data-driven system construction, showcasing Doubao Work's capabilities in browser operation, multi-modal understanding, file processing and application generation, as well as collaborative workflows with Feishu Docs, multidimensional spreadsheets and other products.

Especially in the internal linkage of the ByteDance ecosystem, Doubao Work can be regarded as the intelligent brain of Feishu, while WorkBuddy acts as the neural interface connecting various business systems.

For example, in the real-time inspection of live streaming rooms, Doubao Work can open Douyin live rooms via its built-in browser for screen recording and video understanding analysis, then combine with Feishu Magic Note to transcribe audio content into text, automatically scan screen layout, speech compliance and prohibited terms against the quality inspection standard document, and finally output a complete Feishu document with full verbatim transcript, supporting secondary inspection targeting identified issues.

The moat lies in organizational context

In the AI office track, Tencent's WorkBuddy wins with zero threshold and general agent capabilities, while the entry of Doubao Work means the next stage of competition will focus on who can penetrate deeper into organizations and obtain more enterprise contextual data.

An AI office Agent usually consists of three parts: the bottom layer is the model, the middle layer is Harness, which refers to the full set of engineering infrastructure and operation framework that governs and standardizes model behaviors, and the top layer is contextual data, which determines its level of understanding of an organization and work objectives.

At present, competition at the model layer is ongoing, but it is difficult to create a huge gap. Engineering capabilities at the middle layer are also starting to converge, and connecting different external tools can realize document processing, spreadsheet work and long task execution. Only the top layer of data permissions determines the capability ceiling of an enterprise office Agent. We should not only look at which Agent is smarter, but also see which Agent stands closer to the data.

That is to say, in the first half of the competition, the priority is what AI office can do, while in the second half, what AI office can know is far more important.

For example, writing a weekly report is essentially a summary based on the context of meetings, group chats and document projects of the past week, which should contain data and details. If the Agent can access all your work data and organizational context, it can easily complete the task. If you need to manually provide all documents and projects to it, it is no different from modifying a draft you have already finished.

Therefore, the ability to fully understand enterprise context over a week becomes the key to the real competence of AI office products. But the question is, are enterprises willing to let their internal contextual data be fully accessed by AI?

This is also the experience accumulated by Feishu, DingTalk and WeCom as office products over the years. For example, Doubao Work inherits the existing permission system of the organization, and employees' Agent cannot access data beyond their authorization scope, so the boundary of intelligent collaboration has long been clearly defined. For enterprises, they will dare to hand over contextual data to Agent only when the data is stored inside their own system.

The core is to let AI unleash productivity within the enterprise's security boundary, and the primary premise is that after the Agent is connected, there is no need to build a new independent security architecture, and it can directly reuse the organization's existing permission rules.

Whether it is WorkBuddy with first-mover advantage or the newly released Doubao Work, their current goal is to acquire more enterprise data, enabling the long-accumulated meeting records, document assets and business processes of enterprises to operate on top of AI.

Is AI office a result or an end goal?

At present, WorkBuddy, Doubao Work and Qwen Office are all phased results of AI entering enterprises, rather than the end goal itself.

In the past, the main players in the office track were WeCom, DingTalk and Feishu. Now we have entered the Agent era, users' core demand has shifted to intelligent collaboration, and the core entry point of products has changed from IM and document interfaces to AI agents that can independently disassemble tasks and execute operations. The main players have become WorkBuddy, Doubao Work and Qwen Office, and all business units are bound to serve the Agent business.

The most obvious example is that the Feishu product team was merged into Doubao to form a new Doubao product team, while Feishu's GTM system was merged with the Volcano Engine team to establish the "Creativity Service Platform", which is ByteDance's largest adjustment to its ToB business in the past 5 years.

On the same day when Feishu carried out this adjustment, Tencent's WorkBuddy released a major update, integrating with Tencent Docs and launching the "human-AI dual writing" feature, allowing AI to enter the collaborative editing flow of cloud documents as a virtual collaborator.

But this is far from enough, which is why Tang Daosheng issued the internal letter, pushing for interconnection between WeCom, Tencent Docs, ima, Tencent Meeting, iwiki and other products with WorkBuddy, to truly realize one-stop Agent demands.

In front of the AI strategy, office is only a scenario and closed loop that serves AI. All major tech firms are making adjustments and moving towards their goals, ByteDance has taken a big step forward, while Tencent has taken a small step.

For ByteDance, Feishu is responsible for customer penetration and scenario implementation in the office sector, Doubao is responsible for large model capabilities and product experience, and Volcano Engine is responsible for computing power infrastructure, cloud services and commercial monetization. This three-layer architecture integrates ToB resources, forming a business path of AI closed loop. Doubao Work is one of the landing scenarios of the AI closed loop, and the task flow and feedback data generated by these real office scenarios are the most authentic context for verifying and training Agent execution capabilities.

For Tencent, guarding the AI entry point remains the top priority.

WorkBuddy is already one of the products with the highest strategic priority among Tencent's AI applications, its strategic value is self-evident, and it can even be said to be the best proof to refute the argument that "Tencent is slow in AI development". But to keep the work entry point of the AI era within Tencent's ecosystem, its leading edge needs to expand from the breadth of scale to the depth of scenarios, breaking the inter-departmental barriers between applications, especially the collaboration between WeCom and WorkBuddy.

This trend is deepening. ByteDance has already adjusted its direction, and Tencent's integration of WorkBuddy with WeCom, Tencent Docs or other related products is bound to happen.

This article is from WeChat official account "Blue Hole Business" (ID: value_creation), Author: Zhao Weiwei, published with authorization from 36Kr.