Staff Reporter |
Editor | Wen Shuqi
If you work in a major Chinese city, you’ve probably noticed that WorkBuddy ads are absolutely everywhere right now.
Since May, Tencent’s WorkBuddy advertising blitz has spread from Beijing’s Zhichun Road to Shenzhen’s High-Tech Park, and from subway stations to office buildings and even residential elevators. Open any major app on your phone, and you’re likely to see WorkBuddy staring back at you from the feed.
This is one of the most aggressive saturation-level product launches Tencent has pulled off in years. Behind this intense marketing push is a concentrated deployment of Tencent’s channel, computing power, organizational, and ecosystem resources.
A source from Tencent CSIG told us that WorkBuddy has become one of the highest strategic priority products in Tencent’s AI application portfolio. There’s even talk internally that it could be “the third strategic-level product after QQ and WeChat.” Beyond Pony Ma personally attending product meetings, the project’s requests for computing power, technology, and marketing resources are reportedly approved at lightning speed—described internally as getting a “green light all the way.”
Many Tencent employees believe WorkBuddy has a very high chance of entering Tencent’s “Hall of Fame” this year, one of the company’s highest honors for milestone products.
This heavy investment is already paying off. According to a report from third-party market research firm Analysys, in Q2 2026, WorkBuddy topped the domestic office agent platform rankings with 20.97 million monthly PC visits—outpacing ByteDance’s TRAE (second place) and Alibaba’s QoderWork (third place) combined. We’ve also learned that WorkBuddy’s MAU is currently around 20 million, with DAU in the millions, making it the largest office agent application in China by user scale.
WorkBuddy’s technology stems from CodeBuddy, an AI coding tool launched by Tencent Cloud in 2023. After Claude Cowork was released in January 2026, CodeBuddy lead Wang Shengjie came up with the idea for WorkBuddy based on the existing agent platform. He and a few colleagues built the first version over a single weekend and started testing it internally at Tencent.
Around the same time, OpenClaw went viral in China. Tencent quickly spawned multiple “lobster” products like QClaw, WorkBuddy, and Marvis—what outsiders dubbed Tencent’s “Hundred Lobsters War.” By July this year, Tencent had consolidated QClaw’s product center and some team members into WorkBuddy’s department, signaling that Tencent’s early multi-track exploration was converging and resources were increasingly funneling into the WorkBuddy ecosystem.
A prominent AI blogger told us that Tencent’s ability to rapidly integrate its ecosystem resources and its strength in localizing products are WorkBuddy’s most obvious advantages over other domestic office agents right now.
WorkBuddy is already integrated with Tencent Docs and linked with knowledge and collaboration products like ima, Tencent Meeting, WeCom, and Tencent Lexiang. The blogger mentioned that WorkBuddy comes with a ton of localization-friendly Skills—like QQ Mail, WeChat Official Account Assistant, Xiaohongshu Assistant, Feishu, and Google Workspace—that are basically plug-and-play, and the conversational interface lowers the barrier to entry to nearly zero.
Aggressive pricing and free strategies have also fueled early growth. WorkBuddy uses a hybrid model of memberships, credits, and pay-per-model-call, while lowering the entry barrier with registration bonuses, referral rewards, and free model trials. Compared to Codex’s higher usage costs and Claude Code’s strict account controls, WorkBuddy quickly attracted a huge base of domestic users.
At the same time, Tencent hasn’t locked WorkBuddy into its own proprietary models. Instead, it supports Hunyuan, DeepSeek, GLM, Kimi, and others, letting users pick the model based on task and cost. While this sacrifices exclusivity for Hunyuan, it helps WorkBuddy secure its position as the go-to office entry point.
The high-intensity advertising and low-price strategy are all aimed at grabbing prime position in the early stages of AI office adoption.
In July this year, ByteDance carried out its largest organizational overhaul in recent years for its To B business, merging the Feishu product team with the Doubao product team, and folding Feishu’s sales, marketing, and GTM teams into Volcano Engine to unify sales and delivery of MaaS and SaaS products. Shortly after, Alibaba’s “Qianwen Office” began small-scale testing. According to reports, Alibaba Group CEO Eddie Wu and other top executives have been personally involved in key meetings for Qianwen Office.
Many see these moves from both companies as a counterattack against WorkBuddy. One Tencent insider believes Alibaba and ByteDance originally had more mature office products than Tencent, but that advantage has become a burden—their mindset leans toward adding AI to existing office software, whereas WorkBuddy is newer, has more concentrated resources, and carries less baggage. However, he doesn’t think Tencent has already locked in victory: “Feishu and DingTalk both have very mature enterprise customer service systems. It all comes down to how good the products turn out to be.”
AI office is a battle that every major internet player has to fight. Looking overseas, OpenAI and Anthropic are also rapidly shifting their competitive focus from models and chatbots to office entry points.
Tencent President Martin Lau once noted that internet products mostly process existing information, where content production is largely a fixed cost and delivering information to new users has very low marginal cost. But AI is different—every time you deliver intelligent services to a user, it incurs computing and token costs. Therefore, Tencent can’t simply follow the internet-era logic of chasing DAU and user time first, then figuring out monetization later.
“Finding high-value scenarios is just as important—if not more important—than blindly chasing massive DAU and user time,” Lau said. “This is a critical distinction.”
And office is clearly one of those high-value scenarios. It hits all four criteria: high frequency, real necessity, billability, and execution capability—while also controlling a wide array of software entry points in the process. In other words, whoever controls the agent entry point likely controls the future of software distribution, and may well reshape how enterprise profits are allocated.
On another front, WorkBuddy serves as a feedback loop for Hunyuan. Tencent’s Chief AI Scientist Yao Shunyu frequently emphasizes Co-Design internally—having the model team improve models based on real product failure cases, then having the product team redesign workflows based on the model’s new capabilities. Model training requires real-world data, and the office scenarios WorkBuddy targets are among the most suitable testing grounds.
When Tencent released the Hunyuan 3 (Hy3) large model on July 6, WorkBuddy was the first to integrate it with a limited-time free trial. Two days later, Tencent reported that Hy3’s computing power consumption hit its peak, with afternoon queue rates exceeding 50%. Among WorkBuddy users who actively choose models, over 60% now select Hy3.
Given the current low-price strategy, WorkBuddy’s user growth doesn’t yet fully translate into stable paying demand. Dowson Tong, Senior Executive Vice President of Tencent Group and CEO of CSIG, has said WorkBuddy is still in the investment phase, and no KPI for commercialization targets has been set for the team.
The enterprise segment will be the more important revenue source. We’ve learned that WorkBuddy’s ecosystem team is actively engaging with numerous enterprises to advance procurement and joint development. This year, its commercialization focus is on expanding enterprise customer coverage and building replicable solutions in sectors like government, manufacturing, and internet. In the second half of the year, WorkBuddy aims to further lower the barrier for enterprise adoption, hoping to drive model call volumes, Tencent Cloud resources, and follow-up service revenue.
This battle for the next-generation office entry point is just getting started.
