11/19 2026
357
WorkBuddy's daily active users surpassing 10 million (as of early November) is one of the few genuinely encouraging stories in this year's AI application market. It is not just a traffic milestone; it signals that office Agents are moving out of launch-event demos and into the daily task lists of knowledge workers.
Two years ago, an AI product hitting 10 million DAU would have set the whole industry abuzz. Today, the more interesting question is no longer "another hit product has arrived," but what the repricing of the word "entry point" really means once an office Agent's DAU crosses the 10 million threshold—shifting from which software users open to whom users hand their tasks.
01. The weight of 10 million DAU lies in "how much work gets done." WorkBuddy is not a chatbot, nor is it just another office suite installed on a computer. More accurately, it positions itself as a "general-purpose AI Agent work platform for knowledge workers," capable of planning tasks, invoking tools, executing multi-step work, and delivering results.
This definition means its DAU carries a value fundamentally different from ordinary AI applications: users open it not to ask a question a search engine could answer, but to push an industry report all the way from information retrieval and data cleaning to PPT generation, letting the Agent complete a full cross-application, multi-step chain on their behalf.
Opening a desktop Agent frequently means there is real work to do. What is even more telling is the consumption side. Tencent's Q3 2026 financial report, released on November 18, shows that as of early November, WorkBuddy's average daily Token calls had exceeded 100 trillion, making it the AI application with the highest Token usage among domestic large language models. DAU measures "how many people came," while Token consumption measures "how much work was done after they came."
The latter is the real thermometer of how deeply an Agent product is actually used. Since launching in March, WorkBuddy has cumulatively handled more than 9 billion user task requests. Spread over more than eight months, that means tens of millions of tasks are decomposed, executed, and delivered every day on average.
With 10 million DAU paired with this level of consumption density, it shows that WorkBuddy has moved beyond the "novelty tool" stage and begun to embed itself into the daily routines of some knowledge workers. This is also why the number deserves a positive reading: what it proves is not users' curiosity about AI, but their willingness to hand over real work.
02. Acquiring customers at the free tier, monetizing at the Token tier: a business unlike traditional SaaS
Only after usage density rises is there a basis for discussing the business model. Tencent Holdings (00700.HK) President Martin Lau explained WorkBuddy's business model very directly on the earnings call: acquire customers through the free tier, provide a large number of Tokens in the paid tier, and heavy users keep buying more. "We have even seen users spend as much as RMB 40,000 per month on it just to buy Tokens."
Annualized revenue over the most recent seven days exceeds $500 million—more than twice that of Tencent Meeting, which has been online for seven years; Q3 revenue is more than five times that of Q2. This set of comparisons deserves a moment's thought. Tencent Meeting is one of Tencent's most impressive products of the SaaS era, and it took seven years to reach its position. WorkBuddy took less than a year to surpass it in annualized revenue.
It should be noted that the $500 million is an annualized value converted from revenue over the most recent seven days, not full-fiscal-year revenue, and is not fully equivalent to Tencent Meeting's basis. But the more critical information lies in the structure: traditional SaaS sells "seats," with a fixed cost of a few dozen yuan per employee per month regardless of how much or how little the product is used. WorkBuddy's seats also include credits, but the increment lies in usage: the enterprise flagship edition costs RMB 198 per person per month, with each seat including 2,000 credits per month shared within the team, and top-up packs can be purchased after the limit is exceeded. The more users rely on it for heavy work, the more additional purchases they make, and the thicker the bill.
Martin Lau specifically pointed out that this is "fundamentally different from the past model of Chinese SaaS charging fixed subscription fees." In other words, seats determine who can come in, while usage determines how much revenue one person can contribute: the revenue ceiling of SaaS depends on how many accounts can be sold, while the ceiling of the Token business depends on how much workload users are willing to hand over to the Agent. The fragile side of this logic is also obvious.
The free tier can bring in DAU, but the density of heavy paying users determines revenue quality. On social platforms and in user communities, another usage pattern can also be seen: users praise the product while checking in every day to earn credits, or simply switch the model to a third-party API—for a very direct reason: credits are consumed too quickly.
At an annualized conversion, $500 million is about RMB 3.5 billion. Placed against Tencent's more than RMB 800 billion in annual revenue, it accounts for less than 0.5%; its significance lies mainly in growth rate rather than scale. At the same time, Token costs grow with usage, and the harder users push it, the greater the gross margin pressure—Tencent's Q3 capital expenditure had reached RMB 81.43 billion, up about 527% year on year, while net profit attributable to equity holders under non-IFRS fell 6% year on year, and the overall situation remains in a phase of trading investment for growth.
Whether this business can ultimately work does not depend on whether DAU can double again, but on the race between the conversion rate of the paid tier and inference costs. In other words, 10 million DAU is only the precondition for making the business bigger; the real test is getting people willing to do heavy work to keep paying willingly.
03. The real moat is workflow migration cost
Whether the business model can hold up ultimately comes back to the scenario. One easily underestimated factor enabling WorkBuddy to go from zero to 10 million DAU in more than half a year is Tencent's distribution inertia on PC, and another more fundamental factor is the workflow foundation formed by WeChat and WeCom.
The task starting point for a large number of Chinese knowledge workers is already in WeChat: clients raise requests in chat boxes, bosses assign tasks in group chats, and partners send materials via file transfer. WeChat is a traffic entry point, and even more so the place where work tasks happen.
On October 14 this year, WeCom and WorkBuddy jointly released a new version: after an enterprise opens and allocates seats, employees can directly invoke WorkBuddy from the contacts list or sidebar to handle messages, documents, meetings, and to-dos within the authorized scope. This link has not been connected for long: it is the next growth lever, not the reason for reaching 10 million DAU over the past eight months.
According to QuestMobile data, in September WeCom's PC MAU was 20.302 million, higher than DingTalk's 12.568 million and Feishu's 1.935 million—meaning Tencent does not need to acquire customers from scratch, but only needs to "hand over" an Agent within existing work scenarios.
Data from the enterprise edition also verifies the efficiency of this path from another direction: according to Tencent disclosures, the enterprise edition has been implemented in more than 50 industries, covering 50% of the Fortune China 500 and 60% of the top 100 enterprises, with 60% of the top 100 already becoming its customers. In the To B AI business, channel trust and delivery credibility often determine who can win the first order more than model benchmark scores.
Looking one step further, the reason WorkBuddy's 10 million DAU deserves serious attention is not that Tencent has gained another star product, but that it has validated an industry judgment: competition among office Agents is shifting from "model capability" to "workflow embedding."
Gartner predicts that by 2030, agentic AI may affect as much as $234 billion in enterprise application software spending, accounting for about 20% of enterprise SaaS spending. Once an Agent stably enters an enterprise's workflow, data, operating habits, and collaboration inertia stack together, and migration costs will be far higher than switching to another traditional software product.
This is precisely Tencent's most comfortable and most dangerous position in this competition: comfortable because behind WorkBuddy is an office ecosystem connecting documents, meetings, WeCom, Lexiang, and cloud storage—a scenario density competitors cannot replicate in the short term; dangerous because Alibaba has integrated Qianwen Office into a unified entry point, ByteDance has connected Doubao Work into Feishu, and the pursuers have not stopped.
According to Analysys, among desktop office Agents in September, Qianwen Office and Doubao Work had monthly visits of 46.55 million and 30.39 million respectively, while WorkBuddy had 85.34 million—the lead remains, but it is no longer an order-of-magnitude gap.
The entry-point dividend period will not last forever. Ten million DAU is the first lap after the starting gun, not the finish line. But its positive significance is already clear enough: what WorkBuddy has produced is not just a good-looking number, but proof that office Agents can become a business with revenue, stickiness, and ecosystem depth. As for who will ultimately own this business, it depends on who can get users to hand over more "work"—and to do so with confidence.
- End -