China Joins the "AI That Works for You" Race... Spotlight on Tencent, Alibaba, and ByteDance [Weekend Money]

Chinese AI Market Shifts Competition from Chatbots to Agents
"Platform and Execution Power Now Outweigh Model Quality, Favoring Big Tech"

The global generative AI market's competitive landscape is shifting from "how smart the answers are" to "how much real work the AI can accomplish." In China, the focus of AI competition is rapidly moving from chatbots to agents, and it is being analyzed that tech giants like Tencent, Alibaba, and ByteDance—with their massive user bases and robust payment and transaction infrastructure—are likely to secure mid- to long-term leadership.

ChatGPT Generated Image

According to Shinhan Securities, approximately 24% of applied AI projects in China in the first half of 2026 were agent-related. The monthly visits for major standalone office agents surpassed 60 million in June, up from about 20 million in March—a roughly threefold increase in just one quarter. The number of active users of Tencent's work-oriented agent "WorkBuddy" and Alibaba’s "JVS Claw" has increased by 115.3% and 164.4%, respectively, over the past three months. Unlike traditional chatbots, which simply answer questions, these AI agents perform actual tasks on behalf of users, executing multi-step workflows.

Choi Wonseok, a researcher at Shinhan Securities, commented, "Meta’s recent launch of the AI agent 'Muse' is a symbolic event that illustrates how quickly the axis of competition in the generative AI market is shifting." He added, "The message to the market is that the criterion for AI competition has moved from 'how well it answers' to 'how well it handles a wide range of actual tasks.' It is necessary to pay close attention to recent developments in the Chinese agent market against this backdrop."

One key reason the agent market in China stands out is the presence of vast, pre-built platform ecosystems. Messaging, shopping, maps, delivery, travel, and payment services such as WeChat, Taobao, Alipay, Amap, and Meituan are all concentrated within a handful of giant platforms. Choi pointed out, "If the U.S. tends toward an open, cross-service agent structure that connects different providers through browsers, APIs, and MCPs, China leverages super apps, mini programs, and payment infrastructure to handle the entire chain—search, select, reservation/order, and payment—within a single ecosystem. This makes it easier for Chinese agents to link execution capabilities directly to real-world transactions due to the structure of the industry."

As a result, the competitiveness of big tech companies with large platform user bases is expected to become more prominent in the Chinese agent market. Shinhan Securities identified Tencent, Alibaba, and ByteDance as the firms likely to hold leadership in the Chinese agent market over the mid- to long-term. Rather than selling agents directly as a paid service, it is expected these companies will more likely monetize by boosting transactions and conversion rates on existing platforms through advertising, e-commerce, payments, and cloud/API services.

Choi explained, "While independent AI companies focus on subscription and API usage fees, platform businesses rely less on charging directly for their agents and more on indirect monetization—growing advertising, e-commerce, payments, cloud, APIs, and increased lock-in within existing ecosystems. Chinese platform operators’ agents are more likely to function as interfaces that increase transaction frequency and conversion rates, rather than mere paid AI services."

Tencent is regarded as having secured the strongest user touchpoints in the personal agent space. As of the second quarter, WeChat and Weixin integration had approximately 1.44 billion monthly active users (MAU). Mini programs, search, video, mini shops, and WeChat Pay are all already linked inside WeChat, which means Tencent does not need to promote new standalone AI apps to convert existing users naturally into agent users. Choi noted, "This will be Tencent’s greatest strength and a massive entry barrier for competitors."

Tencent has been gradually testing its personal AI "Xiaowei (小微)" within WeChat since June. In September, it began piloting an "AI-to-AI Social" function, where a user’s Xiaowei could communicate with another user’s Xiaowei to coordinate appointments or intentions, then request final confirmation from the user if necessary. This shows AI evolving from being a personal assistant to becoming an agent within the social network.

Alibaba is focusing on transaction-oriented agents. It is pursuing a structure in which the Qwen App is connected with Taobao, Taobao Instant Commerce, Alipay, Fliggy, and Amap to enable food orders, travel searches and reservations, and in-app payments within a unified AI interface. However, fully autonomous payment—where the AI completes transactions without user intervention—has yet to be realized. Due to requirements for individual service authentication, user approval, and payment security, Shinhan Securities finds that the current feasible model is a "controlled transaction," in which users confirm at the final step. Choi noted, "If agent-driven commerce expands, payment errors, fraud, and responsibility issues may escalate, which means financial regulations could become significant factors determining the speed of commercialization."

ByteDance is working on turning the smartphone itself into an execution environment for agents. The "Doubao Phone Assistant" can invoke smartphone functions and control some apps directly through voice and an AI button. Even apps without an API can be controlled by recognizing the screen and manipulating the graphical user interface (GUI). Thus, it is being analyzed that agents could evolve into a new mobile operating system (OS) layer running atop existing apps.

Choi stated, "There is potential for development beyond simple AI apps, possibly creating a brand new gateway for smartphone traffic." He did caution, "However, since the SAEP system introduced by ByteDance is designed so that third-party apps can directly define the limits of agent automation, it shows that in future mobile agent competition, app permissions and service provider consent—rather than model performance—may be the key bottlenecks."

Among independent AI companies, MiniMax and ZhupuAI are also gaining attention. MiniMax is recognized for optimizing models and productization for agents, while ZhupuAI has strengths in computer-use based on AutoGLM and models for enterprise and developer use. Both companies saw sales increase by 283% and 400% respectively in the first half of this year. However, since agent-specific revenue is not yet disclosed, Choi explained that it is difficult to directly link this growth to agent demand.

There are also challenges facing the Chinese agent market. A key issue is the limited interoperability among large internet platforms. Tencent, Alibaba, and ByteDance all consider user traffic, payment, and transaction data to be core competitive assets, so they have little incentive to fully open their key services to competitor agents.

Accordingly, Choi views it as more likely that multiple agent ecosystems will coexist in China, rather than a single “universal agent” dominating all services. While alliances between some platforms and partial opening of external features may increase, there are expected to be limits on cross-platform agents that fully hand over key transactions and user context to competitors. Choi concluded, "This should also be considered a potential factor for excessive competition and profitability pressures among agent companies over the long term."

This content was produced with the assistance of AI translation services.