Alibaba Merges Three AI Agents for Office Work: A Long-Overdue Validation

It's like renovating your house: you bought three sets of smart lights from different brands, each controllable via phone, but you need three apps. Until one day, you find a universal remote and connect them all. What Alibaba did this time was integrate QoderWork, Wukong, and MuleRun—these three lights—into one switch, called "Qianwen Office."

When I saw this news, my first reaction was: finally, someone is taking "Office Agent" seriously. Alibaba has had many moves in AI office before, but they were like loose parts—QoderWork for code collaboration, Wukong for knowledge management, MuleRun leaning towards automation processes. Each had users and reputation, but users had to switch between three places, breaking the experience. Now merging them essentially answers a core question: Does AI office need a super app, or multiple plugins?

Why "Qianwen Office" Instead of "DingTalk AI Upgrade"

Many people ask, why not just stuff these features into DingTalk? DingTalk is already heavy enough; adding three agents might drive users crazy with the menu bar. Making Qianwen Office independent makes logical sense: it targets the new category of Agent Office, not feature stacking on old tools.

[!tip] Key Judgment: DingTalk is for organizational collaboration, while Qianwen Office is for liberating individual productivity. The former needs to "manage people," the latter to "help people." These two things naturally conflict; forcing them together pleases neither side.

Chen Yusen, the new CEO of DingTalk, is responsible for this product, indicating that Alibaba has made structural cuts—DingTalk continues to serve as the enterprise collaboration base, while Qianwen Office acts as an independent product to capture the C-end and light B-end markets. This strategy resembles the relationship between Feishu and ByteDance's "Feishu Docs," but Qianwen Office focuses more on AI automation rather than document collaboration.

What Chemical Reactions Can Occur After Merging Three Agents?

I carefully looked at the positioning of these three products:

• QoderWork: Focuses on code generation and review, suitable for developers. It already has many programmer users but is too far from ordinary office workers.
• Wukong: The name has a strong Alibaba flavor, doing knowledge management, automatically organizing meeting minutes and document summaries. This function is similar to Notion AI and Feishu Minutes, but Alibaba has advantages in e-commerce and supply chain data.
• MuleRun: Focuses on automation processes, such as auto-filling forms, generating reports, and cross-system data synchronization. This is actually an AI upgrade of RPA (Robotic Process Automation).

After merging, the potential capability map of Qianwen Office can be viewed like this:

Capability Module	Input	Output	Typical Scenario
Code Generation	Natural language description	Executable code	Write a SQL query, make an Excel macro
Knowledge Management	Meeting recordings, documents	Structured summaries, to-dos	Organize weekly reports, extract key info
Process Automation	User instructions	Cross-system operations	Pull data from DingTalk, auto-generate PPT

This combination is quite clever. Code generation is "hardcore productivity," knowledge management is a "general need," and process automation is the "glue." Once connected, users only need to say, "Help me organize the minutes of all client meetings last week, generate a sales follow-up sheet, and sync it to the DingTalk group," triggering a complete Agent workflow. This is deeper than most AI office tools currently on the market.

Where Are the Traffic Opportunities? An Assessment from a Creator's Perspective

As a content creator, I see two huge propagation points for this product.

First, comparative testing. Audiences love seeing "who is smarter." I can take Qianwen Office against WPS AI and Feishu My AI for same-topic tests, e.g., "Help me write next week's operation plan, including data analysis and competitor comparison." Which Agent understands context, which automatically connects multiple steps, and which bugs out—this kind of content naturally attracts traffic. Accounts with 500k followers doing such horizontal reviews can guarantee over 10k likes.

Second, roasting "Frankenstein" failure scenes. Merging means dealing with historical data compatibility issues. If the underlying models of the three agents differ, output styles may be inconsistent, leading users to receive weird replies like "half code, half natural language." Such failure videos spread strongly on TikTok/Douyin, with titles like "Alibaba Qianwen Office Failure Record: Asked it to write a weekly report, got a pile of Python code instead." Any AI product encounters such problems early on, but creators' value lies in amplifying them to force product optimization.

However, regarding propagation quality, I worry about three things:

• The name is too generic. Qianwen Office sounds like a variant of "Qianwen," lacking distinction from "DingTalk Office" or "Feishu Office."
• Lack of unique killer features. If it's just packaging three existing things without a killer feature, it's hard to create "wow" moments in videos.
• Pricing strategy unclear. If charging per Agent call, users feel it's expensive; if free, Alibaba needs to invest massive compute resources.

A Clear Trend Prediction

The success or failure of Qianwen Office depends on whether it can spin up a flywheel of "high-frequency scenario, closed-loop experience, user spontaneous sharing" within 6 months. I predict its most likely breakthrough direction is "automated office processes" rather than "content generation." Because content generation has too many substitutes (ChatGPT, Claude, ERNIE Bot), while automated processes require deep binding with the Alibaba ecosystem (DingTalk, Enterprise Email, Aliyun Drive), which is the moat.

And Chen Yusen, as the new CEO of DingTalk, his core task is not to make DingTalk better, but to

Original Link: https://www.ithome.com/0/979/426.htm