Tencent launched Buddy One on 29 October, alongside a new Tencent Netdisk, and the product is less interesting than the shape it reveals: four previously separate tools pulled behind one account.
WorkBuddy, Tencent Docs, ima and Tencent Design now sit in one suite. Underneath them is a shared base of unified identity, a single credit pool and a single data pool. A user writes a document and the workbench routes it to Tencent Docs, asks for a design and it goes to Tencent Design, needs research and ima handles retrieval and answer.
The routing layer is the product
Each tool does one thing. Tencent Docs handles collaboration and documents. ima is built on a knowledge base and does retrieval, writing and question answering, plus recording-to-minutes, image generation and conversion of documents into mind maps or podcasts. Tencent Design covers visual work and layout.
What Tencent added is the layer that decides who gets the request, so a user stops moving material between applications. That is a smaller claim than it sounds in a launch keynote and a larger one in daily use, because the cost of enterprise AI has never been model quality. It is the friction of assembling output from four places.
Storage is the part with a moat
The more consequential addition is Tencent Netdisk, positioned as unified storage for the whole suite. Its value is not capacity. It is that the context and outputs an agent generates during work get retained and can be called on later, which converts a session into an organisational asset.
The reference point is ByteDance, which folded Doubao Work into Feishu at the end of August so that material produced jointly by staff and AI lands back inside Feishu. Tencent Cloud has run an enterprise AI knowledge base, LearnShare, for longer, drawing on WeCom and Tencent Docs. Buddy One plus Netdisk pushes that capability from an enterprise knowledge product down to an individual’s workbench.
Someone has already priced this
Microsoft Copilot has passed USD 37 billion in annualised revenue, and in June Microsoft added Cowork, billed per use and built to execute tasks across tools rather than within one. That combination, a subscription base plus a metered agent layer, is the model Tencent is now assembling.
The domestic field is crowded in a different way. WPS AI reported 80.13 million monthly active users, about triple a year earlier. Feishu is attacking through knowledge Q&A and enterprise agents, DingTalk through its Wukong platform for business processes. Gartner’s forecast is that more than 80 per cent of enterprises will use generative AI in at least one core process by the end of 2026, and one OFweek estimate puts the average efficiency gain from AI office tools at 64.3 per cent.
The constraint nobody has solved
Tencent’s own account of the data pool points at the hard problem. Qwen’s office product team has calculated that a mid-sized to large enterprise holds roughly 150 petabytes of internal documents, while the best available context window is about one million tokens. Loading everything is not an option, so the agent retrieves from a low-quality pile and hopes.
Making data usable rather than merely stored is the open question, and it is why Adobe rebuilt GenStudio into an agentic content supply chain in April. The industry has concluded that content should be reusable and retrievable rather than disposable.
Tencent’s advantage is that it owns the workbench, the documents and now the storage. Its problem is the one every suite vendor has: bundling raises switching costs without guaranteeing that any single tool is the best available, and enterprise buyers increasingly test tools one at a time.
Editor’s note: This is an adapted translation of the original OFweek report. It has been trimmed and restructured for readability for an international business audience. The full original (in Chinese) is at https://www.ofweek.com/ai/2026-10/ART-201715-8120-30709417.html.