Why does Tencent want WorkBuddy to work across apps?

  • Tencent says WorkBuddy has handled 9 billion user task requests since its launch in March 2026, and its enterprise edition has been adopted by 50% of China’s top 500 companies.
  • Tencent’s management believes the bottleneck to enterprise AI adoption is shifting from model capabilities to how disparate data, applications and business processes can be integrated. The company therefore positions WorkBuddy as a cross-app coordination layer that can call on different tools to complete complex work.
  • Nomura believes competition among workplace agents increasingly depends on the ability to connect proprietary and third-party enterprise applications and complete a range of customer workflows. Alibaba and ByteDance are also strengthening integration among their own products and expanding connections to third-party applications.

Which products has Tencent connected?

  • The newly launched Buddy One connects WorkBuddy, Tencent Docs, Tencent Meeting, ima and Tencent Weiyun, offering a unified account, shared AI quota and bundled subscription. Users can retrieve authorized information across apps through WorkBuddy, reducing the need to switch repeatedly between them.
  • An upgrade on 14 October allows WorkBuddy to run directly within WeCom, read authorized messages, documents and meeting records under a unified enterprise identity and permissions framework, execute tasks and return results. Administrators can also centrally manage permissions and AI usage.
  • Nomura sees these moves as Tencent’s effort to ease barriers between its internal products. A ByteDance workplace-agent business expert had previously expressed concern that barriers between Tencent’s internal departments could affect WorkBuddy’s ability to compete for enterprise customers.

Why do two companies lead different parts of the model services market?

  • IDC data show that revenue in China’s model-as-a-service (MaaS) market reached 10.6 billion yuan in the first half of 2026, expanding nearly 10-fold year on year; token consumption rose 433% year on year. Large language models accounted for 47.0% of revenue and video generation for 32.7%.
  • ByteDance’s Volcano Engine led the overall market with a 45.1% revenue share, while Alibaba had 25.2%. In the large language model services segment, however, Alibaba held 46.6%, ahead of ByteDance’s 10.2%.
  • Based on IDC data, Nomura estimates that ByteDance’s total MaaS revenue in the first half was approximately 4.8 billion yuan, 1.8 times Alibaba’s approximately 2.7 billion yuan. Their large language model revenue was approximately 510 million yuan and 2.3 billion yuan, respectively, with Alibaba’s about 4.6 times ByteDance’s.

How do the two advantages affect the competitive assessment?

  • Nomura believes ByteDance’s overall lead comes mainly from non-large-language-model services such as video generation. The bank estimates that Seedance accounts for more than 80% of China’s video-generation MaaS market and attributes its advantage to model capabilities and ByteDance’s existing short-drama ecosystem. Its discussions with independent production companies indicate that content creators widely use Seedance to make AI short dramas.
  • Nomura believes Alibaba’s advantage in commercializing large language models is supported by the Qwen model ecosystem and its cloud infrastructure. As enterprises adopt AI agents and coding applications, this advantage provides a foundation for future monetization of AI cloud services.
  • Nomura regards ByteDance as Alibaba’s strongest competitor, but believes competition is relatively mild at this stage: the overall market is expanding rapidly, and the two companies focus on video generation and large language models, respectively.