Targeting Alibaba and ByteDance! Report: Tencent is secretly planning WeChat AI agents, may open to all users within the year

Tencent is betting on breakthroughs in the AI race through the WeChat ecosystem. On March 10, The Information reported that four sources revealed Tencent is secretly developing an AI agent for its WeChat app, with the project designated as a high-priority confidential plan, initiated as early as the first half of last year. According to the plan, Tencent intends to launch a grey-box test by mid-year and roll it out to all users in the third quarter. Sources also said that if the feature is not mature, the launch schedule may be adjusted. Once deployed, this AI agent will be integrated with millions of mini-programs on the WeChat platform, covering services such as ride-hailing and food delivery, allowing 1.4 billion monthly active users to complete such tasks autonomously. This move will greatly expand the use cases for AI agents in China and directly challenge Alibaba and ByteDance’s first-mover advantages in this field. Riding the wave of the “lobster boom” AI Agent trend, Tencent is ramping up efforts. In fact, whether it’s laying out its AI Agent strategy or leveraging WeChat’s gateway, Tencent has long made moves behind the scenes; this is just another heavy bet following the momentum. Tencent has launched three AI Agents in succession The AI Agent track has seen persistent rising heat, and Tencent is accelerating its layout with frequent moves. In March, Tencent intensively launched three AI Agent products in one day, targeting three core scenarios: "QClaw" for local personal control (supports remote PC operation in WeChat chat window), "WeCom Bot" for enterprise collaboration, and "WorkBuddy" multi-platform office assistant (seamlessly connects to Feishu, DingTalk, and other mainstream tools). Notably, none of the three products are independent apps; instead, they are directly embedded in high-frequency applications such as WeChat, WeCom, and QQ, relying on existing ecosystems to inject capabilities. This approach reflects Tencent’s strategic shift in this sector: abandoning the traditional path of promoting independent clients and instead leveraging WeChat’s gateway advantage, transforming AI capabilities from “tools that need to be specially opened” into “services natively present in the conversation flow.” By cutting out complex setups and enabling natural language invocation, Tencent aims to seize the core entry point for next-generation human-computer interaction at the critical juncture where AI Agents shift from technical competition to mass adoption. WeChat ecosystem forms the core advantage Tencent’s plan is to embed the AI agent directly into WeChat itself rather than as an independent app, with the key logic being WeChat’s irreplaceable ecosystem scale. According to reports, the agent will appear in users’ chat lists as a conversational interface, completing tasks by invoking mini-programs. However, this strategy also reflects Tencent’s dilemma in its AI layout. Citing two people familiar with Tencent executives’ thinking, Tencent cannot risk immature technology disrupting WeChat’s massive user experience. In May 2024, Tencent launched the independent AI app “Yuanbao,” but market response was flat. According to data from China’s AI product tracking website Aicpb.com, as of February this year, Yuanbao had about 109 million monthly active users, far less than ByteDance’s Doubao at 315 million and Alibaba’s Tongyi at 202 million. In comparison, Alibaba has connected Tongyi with its e-commerce, online travel, maps, and Ant Group payment platforms, allowing users to complete tasks such as grocery shopping and booking flights. ByteDance has upgraded Doubao into an agent capable of handling e-commerce and other tasks. Both companies last month released new-generation AI models claimed to be better suited for complex, multi-step tasks. Model selection presents constraints; self-developed path remains to be validated At the underlying model level, the WeChat team has yet to decide whether to use Tencent’s self-developed “Hunyuan” model. According to reports citing three sources, the Hunyuan model has not yet ranked among the industry’s top tier. Two sources said the WeChat team has tested models from Zhipu, Alibaba, DeepSeek, and other Chinese firms, and is also evaluating WeChat’s own smaller models. However, using external models means longer cycles for integrating and validating WeChat’s internal storage data. In terms of talent, Tencent brought in Yao Shunyu from OpenAI in September last year, appointed him as chief AI scientist, gave him authority to lead Hunyuan model development, and allocated sufficient budget to recruit talent from competitors such as ByteDance. Meanwhile, the WeChat team led by founder Zhang Xiaolong is also advancing independent AI model development, releasing two technical papers on the official blog in January on improving model capabilities with limited resources and post-training methods. WeChat tech lead Zhou Hao (Harvey Zhou) reports to Zhang Xiaolong and manages the AI team. Competitive landscape intensifies; Tencent faces catch-up pressure From a more macro perspective, Tencent’s recent push in the AI Agent track is a microcosm of global tech giants battling for the AI assistant gateway. From Silicon Valley to China, major companies are racing to launch AI assistants capable of autonomously completing complex tasks such as programming and shopping, trying to seize the high ground for next-generation human-computer interaction. For Tencent, there are both advantages and pressures in this race. Compared to the more proactive moves of Alibaba and ByteDance in AI, Tencent’s earlier steps were somewhat slower. However, since mini-programs launched in 2017, WeChat has built a major ecological barrier for super apps through its exceptional user experience, inspiring global players like Microsoft, Snapchat, and Google. This ecosystem, accumulated over eight years, should be Tencent’s core advantage in the AI agent era. The key question is: how to effectively convert this ecological advantage into AI competitiveness—surpassing rivals—without disrupting the existing user experience. The real battlefield: Entry point competition On the surface, this battle is about whose “agent” is more powerful and easier to use. But fundamentally, the focus is shifting from “features” to “entry point”—the usage logic of AI is being reshaped. In the past, users had to manually open independent apps, ask questions, wait for generation, and then manually transfer results; this process was too long, raising hurdles and hindering mass adoption. Now, from QClaw’s integration with WeChat, WorkBuddy’s compatibility with Feishu and DingTalk, ArkClaw’s deep adaptation with Feishu, to CoPaw connecting DingTalk and Feishu, these products collectively signal: The main battlefield of AI agents is migrating from independent apps to communication and office tools where people are already active. Whoever is closer to users has a more valuable gateway. When AI truly lives in your chat list, “using AI” finally becomes part of ordinary life. Meanwhile, an unsolved deep issue emerges: Security. When AI agents are granted the ability to control local devices, access files, and execute sensitive commands, how to prevent unauthorized access, defend against malicious prompt injection, and ensure data privacy remains challenging. While some enterprise solutions have addressed parts of this, a universal “security baseline” for the industry is still a long way off. From “three hours unable to install” to “download and use, send a message and get tasks done,” the speed of AI agent adoption is exceeding everyone’s expectations. But the battle for entry points has only just begun. Risk Disclosure and Disclaimer The market entails risks; invest with caution. This article does not constitute personal investment advice and has not taken into account individual users’ specific investment objectives, financial circumstances, or needs. Users should consider whether any opinions, views, or conclusions in this article are suitable for their circumstances. Investing based on this is at your own risk.