By Hamza Faisal
ByteDance has begun winding down Coze Enterprise Edition, its agent-development platform for businesses, in the latest sign that China's tech giants are abandoning the scattergun AI product strategy of the past two years. Under a shutdown notice issued by the company, the platform's Enterprise Standard and Flagship editions will stop accepting new customer purchases from November 30, 2026, with existing packages fully terminated by December 31, 2027 at the latest and unused portions automatically refunded.
Coze Enterprise Edition launched in the second half of 2025, meaning its commercial life will have lasted barely a year. The move follows Tencent's decision to shut down Yuanqi, its own no-code agent platform, which stops service on November 9, 2026. Two similarly positioned agent builders from China's two largest tech companies exiting within weeks of each other marks a clear turning point for the industry.
A graveyard that grew fast
According to a tally by Tech Planet, Alibaba, ByteDance, Tencent, Baidu and several startups have shut down or merged 38 AI products, with 25 of them disappearing in the second half of 2026 alone. Tencent has been the most aggressive in pruning: its Translator Jun, live for nearly a decade, went offline in March 2025 with its capabilities folded into Yuanbao, and its QClaw desktop assistant, launched in March 2026, will stop operating on December 24, with users directed to WorkBuddy.
Yuanqi's wind-down shows how abrupt these exits can be for users. Tencent announced the shutdown on October 8, stopped new agent creation and API key distribution on October 29, and will switch off application access, data read and write, API calls and the developer console on November 9. Agents published to WeChat channels will stop responding, and user data will be deleted after January 9, 2027, apart from records Tencent must retain by law.
Why the giants are narrowing their focus
Three pressures are converging on the standalone AI app. First, the models themselves keep absorbing single-purpose tools: a feature that once justified its own product becomes a prompt inside a flagship chatbot. Second, compliance rules are tightening, including around anthropomorphic interaction, raising the cost of running many small products. Third, computing costs remain high, and spreading inference spend across dozens of experiments is hard to justify once the novelty fades.
The result is consolidation around a handful of core platforms: ByteDance's Doubao, Alibaba's Qwen and Tencent's WorkBuddy. In July, Tencent moved QClaw-related operations and part of its team into the WorkBuddy division, and in June and July, Yuanbao, Doubao and Tongyi Qianwen each removed or tightened user-built agent features on the same day. The giants are no longer casting a wide net; they are picking their entries and defending them.
Startups feel the squeeze hardest
When a giant retires a product, its technology, users and team can usually be reassigned inside the group. Startups rarely have that luxury. Moonshot AI's overseas character-chat product Ohai and its audio-video generation tool Noisee stopped iterating in September 2024, with resources flowing back to its Kimi chatbot, and StepFun's role-play agent Maopaoya suffered a similar fate. For smaller builders, a discontinued product line often means the end of a commercial bet entirely.
For businesses that built on Coze Enterprise Edition, the practical lesson is the one Yuanqi users learned last month: export early, and do not assume a platform will outlive the project built on it. ByteDance gives customers until the end of 2027, a longer runway than Tencent offered, but the direction is unmistakable. China's AI race is entering a consolidation phase, and the standalone app boom is its first casualty.



