When Tencent Holdings founder Pony Ma described the company’s AI effort as a leaky ship this past summer, it was a rare public admission of weakness from a company that had dominated Chinese tech for years through WeChat. The comment landed at a moment when Tencent was visibly struggling to keep pace with rivals like DeepSeek in the post-ChatGPT race to build dominant foundation models.
The response came from Dowson Tong, one of Ma’s key lieutenants and the executive leading Tencent’s cloud business. Rather than pouring more resources into chasing competitors on raw model performance, Tong’s answer was to retool the company’s AI strategy around enterprise use cases instead, a bet that Tencent’s actual advantage lay less in building the best foundation model and more in wrapping AI around the business tools companies already used every day.
The centerpiece of that pivot is WorkBuddy, a tool that started as a side project within Tencent’s sprawling internet portfolio before becoming the focal point of the turnaround. WorkBuddy combines Tencent’s existing messaging and office software with AI agents capable of pulling data and generating reports directly, essentially a Chinese answer to the Claude Cowork style of enterprise AI assistant rather than a consumer-facing chatbot competing head-on with ChatGPT or DeepSeek.
The early numbers suggest the bet is paying off faster than a typical enterprise software pivot. By fall, WorkBuddy had reached 32 million monthly active users, a figure Tencent says exceeds the combined user base of its rivals in the same category, a notable result for a tool that was treated as a secondary project only months earlier.
The strategic logic behind the shift is straightforward: rather than competing directly with DeepSeek and other Chinese AI labs on foundation model quality, a race where Tencent had visibly fallen behind, Tong is betting that distribution through Tencent's existing enterprise and messaging infrastructure is a harder advantage for rivals to replicate than raw model performance alone.
That distinction, consumer chatbots versus enterprise agents built on existing infrastructure, is becoming a broader fault line in how Chinese tech giants are positioning themselves in AI. Tencent's pivot under Tong suggests the company has concluded it doesn't need to win the foundation-model race outright if it can win the enterprise-integration race instead, patching the leaky ship Ma described not by plugging every hole at once, but by redirecting where the company is sailing.
