Chen Yusen did not follow the old path taken by Alibaba.

Chen Yusen has served as CEO of DingTalk for 46 days.

The first product has been released.

On July 27, the official website of Qwen Office went online, with Windows and macOS Beta versions launched simultaneously. Many people seeing this news may think it is just an Agent tool.

To be honest, I think instead that this is the most profound change Alibaba has made in recent years.

Because you can barely find the grand, high-profile narrative of the past on this product, what you see everywhere is pragmatism, restraint and minimalism.

The young Chen Yusen did not choose to follow Alibaba's old path.

I. This is not at all the Alibaba we used to know

Let's look back at what Alibaba has done in the past few years.

Taobao, Alipay, DingTalk, Quark, Tongyi Qianwen, IM, enterprise platforms......

Almost every time the era changes, Alibaba wants to build a new entry point from scratch.

The three words "platform", "infrastructure" and "operating system" are almost engraved into Alibaba's corporate DNA.

Therefore, when Alibaba develops AI office products, it always talks about vague grand concepts such as the future office entry point, operating system, and super entry point.

QoderWork, Wukong, MuleRun: three lines of AI Agent projects were run at the same time, with three teams competing against each other under different business groups, each with their own ambitions.

This is so typical of Alibaba. It wants to do everything and give up nothing.

QoderWork develops desktop agents, focusing on code sandbox and system-level control. Wukong develops enterprise collaboration Agent, tied to DingTalk's account system and approval workflow. MuleRun develops execution engine, targeting automated scheduling for the overseas market.

Users cannot tell the differences between them clearly, and I am afraid even Alibaba's internal teams cannot make that clear.

In March 2025, Wu Zha returned to DingTalk, calling for breaking DingTalk down and rebuilding it with AI. Wukong was positioned as an AI-native working platform, which was supposed to rewrite the underlying code of DingTalk and make IM, documents and approval all CLI-based.

What was the result then?

Wu Zha left his position 437 days later. The 75,000-word article *Inside DingTalk* went viral across the internet. The Partner Committee released a statement to draw a line with the old plan within 24 hours. Wukong was reorganized into the new system. Chen Yusen took over the post and started the integration in the first week. The three Agent production lines were merged.

The newly launched Qwen Office has clearly stated on its official website:

IM writing, Office content generation and editing, multimodal content understanding, full-stack web page generation, data source access, and direct packaging of custom skills.

The official website page only has a simple product introduction.

None of the empty concepts such as "platform", "redefine" and "operating system" appear on it.

II. Why did Alibaba change suddenly?

What is really worth discussing is not what Chen Yusen did in these 46 days, but the overall transformation of Alibaba.

In the past two years or more, a series of changes have taken place in this company. Wu Yongming took over as CEO, the "1+6+N" strategy was completely abandoned. Non-core assets were divested one after another. The number of partners was reduced from 26 to 17.

Many people think this is just a personnel adjustment.

In my opinion, this is a forced tactical contraction.

Before 2023, Zhang Yong promoted the "1+6+N" spin-off at Alibaba, hoping that each sub-business could go public independently. Taotian Group, Alibaba Cloud, Cainiao, Freshippo, International Business, Local Services were all separated out, queuing up to ring the bell for IPO.

The capital market did not buy it at all. The IPO plans of Freshippo, Cainiao and Alibaba Cloud were suspended. Some media even reported that internal collaboration broke down: it was said that Taotian and Ele.me needed to re-negotiate the pricing of delivery agreements, Alibaba Cloud charged its sister companies higher fees than external clients, and employees even needed to resign first then re-employ if they wanted to transfer across business lines.

At the same time, Pinduoduo rose rapidly and once surpassed Alibaba in market value. Even Charlie Munger publicly stated: Investing in Alibaba was one of the worst mistakes I had ever made.

To put it bluntly, Alibaba spent three years splitting itself into six parts. Every part was losing money, and none of them operated as an independent viable entity.

After Wu Yongming took over as CEO.

The first thing he did was not expansion, but taking back control. Selling assets, cutting redundant businesses, and concentrating all resources on two lines: e-commerce and AI + cloud.

Alibaba invested 380 billion yuan in AI and cloud infrastructure, a figure that directly exceeded the total sum of the past ten years.

In 2025, Alibaba fought three battles: the peak daily order volume of Taobao Flash Purchase reached 120 million, Amap's Street Scan List cut into the local life service market, and the downloads of Qwen App exceeded 10 million within one week of public beta. The share price of Hong Kong stocks rebounded sharply for the whole year, with an annual increase of nearly double, and the market value once returned to the HK$3 trillion mark.

Until mid-2026, DingTalk changed its top leadership, the three Agent production lines were merged, and Qwen Office was officially launched.

From addition to subtraction, from dispersion to focus.

Qwen Office itself may not be a groundbreaking product, but the signal it sends is extremely significant: Alibaba has finally abandoned the old Alibaba-style all-encompassing approach.

Rather than trying to define a new species, it is better to step back, and focus on making the current tool well in a down-to-earth manner.

III. Replacing Wu Zha: From building platforms to making tools

Let's look at Chen Yusen from another perspective.

Born in 1992, he founded Chaitin Tech, a cybersecurity company, at the age of 22, which was later acquired by Alibaba Cloud. In 2025, he started an internal entrepreneurship project at Alibaba Cloud and developed MuleRun, an Agent execution engine targeting the overseas market.

He is not a platform builder, he is a product developer. More accurately, he is a tool maker.

Who is Wu Zha? The founder of DingTalk. His underlying logic is to break down the existing system and rebuild it, to forcefully create a new office entry point. This is the typical Alibaba gene: top-down, to change the world, to rewrite the rules.

Wu Yongming replaced Wu Zha and appointed Chen Yusen, which itself is a highly targeted product judgment. Alibaba has completely abandoned the fantasy of controlling the entry point in AI office business, and turned to the tool-oriented mindset.

This means that from the day of the appointment, the product gene of Qwen Office has been determined.

What Chen Yusen did after taking over DingTalk is essentially Tencent-style "product cleanup", cutting off the empty and unrealistic grand concepts of platforms, and delivering a minimalist user experience.

In the first week after taking office, Chen Yusen promoted the integration of Wukong and MuleRun. In early July, he further brought QoderWork into the unified system.

The official website was launched within 46 days.

The extremely fast speed only proves one thing: once the direction is clear, there will be no more internal tug-of-war. Previously, Alibaba struggled on AI office business for more than two years, and could never reach a consistent internal consensus on whether to make a platform or a tool, to build an entry point or to provide capabilities.

Now the consensus has been made: just make tools.

The pricing of Qwen Office also reflects this attitude. The free version costs 0 yuan. The personal standard version is priced at 78 yuan per month for a limited time, and the premium version is 158 yuan per month.

There is no fancy rhetoric about enterprise digital transformation solutions, it is just a tool pricing logic: you pay for the functions you use.

IV. Everyone is stepping back

In fact, Alibaba's new mindset is also the result of cruel market lessons, because it has found that the path of forcibly transforming user habits is a dead end.

Around the world, all the truly successful AI products choose to "step back".

Google integrated Gemini into Search, Gmail, Docs and YouTube. At the earnings call for the second quarter of fiscal 2026, Google stated that the monthly active users of AI assistant Gemini now exceed 950 million. Its growth follows only one logic: embed AI into the old products that billions of people are already using.

Microsoft embedded Copilot into Windows, Edge, Word, Excel, Outlook and Teams. Instead of building a completely new system from scratch, it just added an AI colleague to the old Office suite.

Apple announced its cooperation with Google at WWDC 2026, rebuilding Siri based on Gemini. Siri has become an independent App that can access photos, emails, messages and calendars. Users already have iPhones in their hands, so the AI just runs inside the existing devices.

Domestic desktop agents are also competing to seamlessly embed into the existing workflow of enterprises.

All global tech giants are doing the same thing: do not fight against users' old habits, go to the place where users already are, and get the work done.

Why is that?

For Alibaba, the overbearing logic of forcibly transforming users represented by DingTalk in the past no longer works.

According to the long review article written by the project witness of the *Inside DingTalk* project, the DingTalk ONE project tried to change the mode of "people looking for tasks" to "tasks looking for people", to let AI reorganize all office information flows; the project invested more than 1 billion yuan in training costs, analyzed 1850 user demands in four months, and rewrote more than 20 product lines.

After its launch, the next-day retention rate plummeted directly from 45% to 18%, and the 7-day retention rate fell below 10%. More than 60% of users uninstalled it after opening it three times.

Users directly complained that this AI was a tool for managers to monitor employees at work. Negative reviews flooded Xiaohongshu, Zhihu and Weibo, with keywords all pointing to surveillance, involution and oppression.

A product that tried to redefine office work was voted down by users with their feet.

The most ironic thing is that ONE found the correct answer before it was shut down: when the team abandoned the all-encompassing approach and only made a simple judgment that an empty homepage meant all urgent tasks had been completed, the retention rate rose from 10% to over 30%.

But it was obviously too late.

Users will not change their entire working method just for AI. Even if you develop an AI operating system, users will still use the Office on their computers. No matter how many times you shout "redefine", users' ten-year working habits cannot be changed in a single day.

V. Platforms step back, get close to users

If you look deeper, the significance of this incident is far greater than Alibaba itself.

Essentially, Alibaba has understood the drastic change of business philosophy in the AI era: the internet is moving back from the platform era to the tool era.

Over the past 20 years of the internet, platforms have grown bigger and bigger. Taobao, WeChat, Douyin, DingTalk and Feishu all want to become platforms. Every company wants to become an operating-system-level existence.

After the emergence of AI, the logic is completely reversed. Platforms have to step back behind the scenes, and what really stands at the front end are the tools that solve specific tasks one by one.

The platform still exists, but it starts to become invisible.

When users turn on their computers, their mindset has changed. Previously, they thought about which App to use today. Now they think about letting the AI help them finish the task.

As for whether it is Qwen, Doubao or Hunyuan running behind the scene, users do not care at all. Whoever can get the task done will be used by users.

The current competition is no longer about who owns the users, but about who is closest to the users.

Alibaba believed in owning users for the past 20 years, building platforms, creating entry points, and locking in user groups. The launch of Qwen Office shows that this company has finally recognized the reality in the AI era.

You can no longer lock users in, you have to go to their side.

This transformation is more profound than replacing any CEO.

Words beyond the page:

Many people think that the competition in the AI era lies in the size of the model, but I increasingly believe that it lies in business restraint.

The more you want to reinvent the world, the more likely you are to lose in the end.

This time, Chen Yusen did not follow Alibaba's old path, and he represents that Alibaba has finally seen the cruel truth of the AI era.

The products that can survive in the end are often the ones that quietly do one thing well.

This article is from WeChat Official Account "Beyond the Page", author: Huahua, published with authorization from 36Kr.