China’s AI Office-Agent Race: Tencent’s Lead, Alibaba’s Reset, and 360’s SMB Bet

China’s AI office-agent market is becoming a contest for the workplace’s default interface. The leaders are not simply trying to produce better answers: they are trying to place AI inside communications, documents, calendars, task flows, and enterprise data.

Tencent, Alibaba, and 360 entered the late-July product cycle with notably different playbooks. Tencent has scale and workplace integration; Alibaba is simplifying a fragmented product lineup around DingTalk; and 360 is using generous trials to win smaller organizations. The common goal is to turn an AI assistant into a repeat-use work surface.

Tencent: convert desktop reach into workflow ownership

Tencent’s WorkBuddy recorded 20.97 million monthly PC visits in June 2026, ranking first among China’s domestic PC-side AI-native office agents. That was nearly 2.4 times its 8.85 million visits in the month of its March release, according to Analysys data reported by 36Kr. 6

The significance is less the traffic figure alone than Tencent’s potential distribution path. WorkBuddy sits alongside Tencent’s workplace and collaboration products, while WeCom’s internally tested AI assistant, Dayuan, is designed to work from existing enterprise context. Dayuan can access Enterprise WeChat documents, spreadsheets, meetings, and calendars; its reported functions include summarizing unread messages, drafting weekly reports from work records, and tracking to-do items. 38

That is the strongest version of the ecosystem strategy: the agent does not ask workers to move into a separate AI destination. It appears inside the systems where coordination already occurs.

Alibaba: consolidate the office-agent entry point around DingTalk

Alibaba’s response has been to reduce product sprawl. Qianwen Office brought together QoderWork, Wukong, and MuleRun under Chen Yusen, the new CEO of DingTalk. Its website went live on July 27 with Windows, macOS beta, and HarmonyOS computer-preview versions; 36Kr later reported that Qwen Office entered public beta on August 3. 76

The consolidation matters because an office agent depends on a coherent product surface. Rather than sending users toward multiple specialized tools, Alibaba is building a single Qwen-powered office product that can ultimately be embedded in DingTalk’s workplace environment.

Alibaba’s challenge is execution after the reorganization: turning a unified product into frequent, trusted use across documents, communication, and organizational processes. The advantage is not merely access to a model; it is the possibility of connecting model capabilities to an established enterprise workflow system.

360: use free capacity and hands-on adoption to reach smaller firms

360 launched Nano Work on July 28 as an enterprise intelligent-agent work platform aimed at business owners, entrepreneurs, sole proprietorships, and employees. Early users were offered a 100-million-token trial allocation. 3234

This is a customer-acquisition strategy as much as a product launch. A large trial allowance reduces the cost of experimentation for businesses that may not yet have formal AI budgets or in-house implementation teams. China Daily reported that 360 paired the offer with training and hands-on support, initially aiming to help 1,000 small businesses deploy AI and improve operations. 34

360 does not need to replicate Tencent’s communications footprint or Alibaba’s DingTalk position to compete. Its plausible opening is a practical, lower-friction proposition for smaller organizations: help users move from occasional AI use to repeatable agent-supported work.

Why distribution and embedded context matter more than a feature list

The market is already expanding quickly. Analysys data cited by 36Kr put monthly visits across 17 mainstream domestic desktop AI-native office agents at more than 60 million in June, up from about 20 million in March. 4

But visits do not establish a lasting winner. For an office agent to complete useful work reliably, it needs appropriate access to information and tools: messages, documents, meetings, calendars, task systems, and organizational context. Platforms that already host those activities have a structural advantage because they can make the agent feel like part of daily work rather than another tab to open.

That does not guarantee an oligopoly, but it raises the bar for standalone entrants. Smaller vendors may still compete through specialized workflows, deployment flexibility, security positioning, or lower-cost adoption. The broader direction, however, favors companies that combine AI capability with an existing workplace ecosystem.

ByteDance’s monetization question

ByteDance remains relevant because of the connection between its AI products and Feishu, its workplace collaboration platform. In this market, a move toward paid office-agent offerings would make the central commercial question explicit: can an agent deliver enough dependable, measurable value that customers will pay on a recurring basis?

That question applies to every competitor. Free tokens can encourage trials, and monthly visits can demonstrate interest, but neither alone proves a sustainable business. Subscription conversion will depend on whether teams see agents saving time or completing valuable work in routine processes.

What to watch next

The most useful indicators will be operational rather than promotional:

• Recurring workflow use: whether agents are repeatedly used for meeting follow-up, reporting, customer-service preparation, sales coordination, and approvals.
• Depth of integration: whether an agent can work safely across messages, files, calendars, and task systems without forcing people to copy and paste context.
• Trust and control: whether businesses can manage permissions, review outputs, and rely on agent actions in sensitive work.
• Monetization: whether free trials and broad access convert into paid seats or outcome-based spending.

The prize is large, though forecasts should be treated cautiously. iiMedia Research projected China’s broader AI-agent market to rise from 80.4 billion yuan in 2025 to 696.8 billion yuan by 2030. That estimate covers the broader AI-agent market, not desktop office agents alone. 9

For Tencent, Alibaba, 360, and ByteDance, the decisive metric will be whether their products become indispensable parts of business operations. The company that owns the routine workflow—not just the AI conversation—will have the stronger position.