Tencent officially launched the WorkBuddy open platform on September 2, bringing in more than 100 partners across smart hardware, industry applications, and developer services. The platform exposes an Agent Harness that packages model calls, context management, task planning, tool use, and execution feedback, allowing companies to connect existing expertise and data instead of building every Agent component themselves. The article says WorkBuddy had already accumulated a user base: Analysys reported more than 20 million monthly PC visits in China’s AI-native office-agent market in June, while Tencent said the product had reached more than 50 industries since its launch in early March. Its ecosystem includes over 30 hardware brands and more than 30 software or service partners, with examples spanning finance, CRM, legal, healthcare, education, and design. Tencent said SalesYi’s three-person team built an associated Agent in two months, and FanRuan MOSS recorded more than 10,000 calls in its first week. The analysis argues that opening the platform creates a central AI entry point where users state goals and Agents select the relevant data, tools, and specialist capabilities. Unlike traditional software ecosystems, however, every additional Agent run creates inference expenditure. WorkBuddy may initially be subsidizing the Token bills of its partners in exchange for ecosystem density and usage data, but the article notes that this model becomes difficult if enterprise subscriptions, usage charges, or revenue sharing do not grow as fast as consumption. It compares WorkBuddy with Doubao Work, Qwen Office, and Baidu Dazi: ByteDance and Alibaba are consolidating their own office and enterprise capabilities, while Baidu uses personal and enterprise subscriptions, including seat-based and usage-based charges. The central business problem is to make higher usage produce higher revenue: a healthy ecosystem would let users pay for completed work, partners monetize their capabilities, and the platform share in the resulting revenue. The article therefore proposes tracking users, Agent calls, Token consumption per call, and commercial conversion together. Its conclusion is that ecosystem scale alone is insufficient; each recurring Agent invocation needs a sustainable source of income.
