437 days.

This is the length of time Chen Hang has been in charge of DingTalk for the second time.

From March 31, 2025, when Alibaba announced the acquisition of Two Hydrogen and One Oxygen, marking the founder's return to the helm after four years, to June 11 of this year, when he stepped down as CEO, a total of 437 days passed.

In those 437 days, DingTalk held two important product launches and unveiled "Wukong," the world's first enterprise-level AI-native work platform. It also published two viral resignation articles, received unprecedentedly harsh criticism from Alibaba partners in 27 years, and welcomed its successor, Chen Yusen, born in 1992.

Three questions string together the entire story of 437 days:

What did Chen Hang do wrong? What price did Alibaba pay for this comeback? When Chen Yusen took over in 1992, what was left on the DingTalk chessboard?

The story may begin in 2014 with an apartment building called Lakeside Garden.

1. Return to Lakeside Garden

In 2014, in an old apartment at No. 176 Wenyi West Road, Xihu District, Hangzhou, six Alibaba employees held a weekly meeting around a table.

They just failed once. The previous product was called "Laiwang"—Alibaba once invested 1 billion yuan, launched a massive manpower campaign, and required each employee to attract 100 external users per month. Jack Ma personally endorsed it, and Liu Chuanzhi, Shi Yuzhu, and Jet Li were invited to join, but in the end, it did not shake WeChat in the slightest.

After that failure, Chen Hang (nicknamed "Wuzhao") and a few others "dive" into this apartment again—Lakeside Garden. This place has successively incubated Alipay, Tmall, and Cainiao, making it a lucky place for Alibaba.

The interior of the replica of the lakeside garden in Alibaba's Xixi office area. Image source: Jiazi Guangnian.

In January 2015, the first generation of DingTalk was launched. Its core insight was to deduce functions from the needs of business owners. For example, the "Ding" function, read/unread status, corporate address book, and approval functions—features that were later criticized by many as "strong management and strong control"—answered the most basic and anxious questions in Chinese companies at the time: Did the other party actually see what I said? Did the tasks I assigned actually move forward?

This insight is clearly a sharp breakthrough. DingTalk surpassed 100 million users in its first year and 300 million in three years.

In his DingTalk job posting, Chen Hang referred to his team as a "madhouse," and his T-shirts bore the slogan "BE CRAZY." He himself worked more than 15 hours a day, often from 8 a.m. until the early hours of the morning.

His statement at the mobilization meeting in 2018, "I don't know what you're doing if you go home before 10 o'clock," is still a joke that is repeatedly quoted in the internet industry.

In 2020, things changed. Alibaba announced its "Cloud-DingTalk Integration" strategy, upgrading DingTalk to the "Big DingTalk Business Unit" and integrating it into Alibaba Cloud Intelligence Business Group, directly managed by Zhang Jianfeng, then president of Alibaba Cloud Intelligence. Chen Hang's DingTalk was required to deeply integrate with Alibaba Cloud's private cloud architecture, and the originally independent standardized product roadmap began to clash with the customized needs of large customers.

In 2021, Chen Hang was removed from his position as CEO of DingTalk and became an assistant to Alibaba Group CEO Zhang Yong. Ye Jun succeeded him as CEO of DingTalk . In the same year, Chen Hang officially resigned and founded "Two Hydrogens and One Oxygen" (HHO) to expand into cross-border markets, launching some small smart hardware products.

During his years at HHO, he wasn't highly regarded by outsiders. A startup with no shortage of stories but lacking funding failed to achieve any truly impressive results. He later told the media that during his time at HHO, he "learned to strictly control costs, stipulating that only one out of three elevators could be in operation." These are the kinds of trivial matters that executives at large companies don't need to worry about.

However, in the years since he left DingTalk, DingTalk's situation has become increasingly awkward.

In 2024, it remained the largest enterprise office application in China: 700 million users and 26 million enterprise organizations, but its commercialization progress was overtaken by Lark. During Ye Jun's era, DingTalk allocated 60% of its R&D resources to customized solutions for large clients, leading to complaints from the ISV ecosystem that "the platform has become a channel provider."

In February 2025, Alibaba announced a three-year AI development plan, investing 380 billion yuan. That same month, during an earnings call, CEO Wu Yongming clearly positioned DingTalk as "Alibaba's most important AI application for the B2B sector."

One detail is worth noting: Wu Yongming and Chen Hang's connection dates back to 1999. That year, Chen Hang first joined Alibaba as an intern, and his mentor was none other than Wu Yongming. Later, he left Alibaba twice, and returned twice at Wu Yongming's invitation.

On the evening of March 31, 2025, news broke that Alibaba had acquired shares from HHO investors. Following the transaction, Chen Hang became CEO of DingTalk.

The story of the discerning eye and the talented individual has been told countless times within the Alibaba ecosystem. But few people question: when a founder is invited back to the company they created, are they there to put out a fire or to rewrite the script?

2. The "tightening spell" of April

Last April, just a few days after Chen Hang returned, DingTalk was in an uproar.

A series of measures have been revealed: clocking in at 9 a.m., shortening lunch break, requiring employees to be in work mode by 1:15 p.m., evening summaries, banning social media apps such as WeChat, Weibo, and Xiaohongshu, and standardizing external communication language as "Sorry, I only have DingTalk".

Even more stringent measures include: tech teams being required to check their code volume, and programmers who have written zero code in the past three months will be laid off; even all management positions are required to learn Python in order to reduce purely management roles; and product managers are required to visit three companies each week for co-creation.

Chen Hang himself patrols the building every night at 10 p.m., giving thumbs up to students who work overtime.

On social media platforms at the time, "DingTalk refugees" became a new label. Some employees changed their bios to "DingTalk is starting to get busy" and asked competitors if they were hiring. Someone joked in the anonymous section: "I got rejected on a blind date after adding them on WeChat, and they just left when they heard I could only use DingTalk."

The author of the 75,000-word article "Being Inside DingTalk," which went viral on June 4, is Teng Yaxin (nickname "Yousu"), the core product manager of the DingTalk ONE project. In the article, she recounted her experience of being subjected to a "pledge of allegiance" style of questioning during her DingTalk interview.

Chen Hang repeatedly pressed, "Why can't we do it? Is there anyone left in your father's family? Is there anyone left in your mother's family? Are your maternal grandparents still alive? Is it really impossible to find them? Is it really impossible to gather six family members who can log into DingTalk?"

Although these questions seem absurd, Chen Hang used this logic to screen like-minded people when he incubated DingTalk in 2014; and in 2025, candidates' first reaction to this interview culture at DingTalk was PUA (Pick-Up Artist) tactics.

The same action has a different meaning depending on the object of the action.

At that time, Chen Hang also launched a "ground-based movement": requiring each member of the product, R&D, and operations teams to serve as customer service for two hours every day.

This action was later proven to be the key to Chen Hang uncovering DingTalk's "data illusion." Previously, the customer service team reported that "the rate of transferring to human agents was only 15%, and all reviews were five-star ratings." However, Chen Hang's on-site visits revealed that many users complained about "irrelevant answers" when they inquired, "no response to requests submitted for over a year," and "inability to find a human agent."

Chen Hang quickly updated the data: DingTalk's actual customer satisfaction rate is only 30%.

3. AI DingTalk 1.0, DingTalk ONE is born

The iron-fisted Chen Hang quickly produced the results.

On August 25th last year, DingTalk held its 10th anniversary press conference, which was also the launch of AI DingTalk 1.0.

At the event, Chen Hang launched five products in one go:

DingTalk ONE: A new generation of interactive portal that can automatically identify "the most important thing at the moment".
DingTalk A1: A 3.8mm thick AI recording card that magnetically attaches to the back of your phone and is powered by a Hengxuan 6nm chip.
AI Forms: Enables businesses to generate AI applications with zero code.
AI-powered transcription: Supports cross-language meeting recording in 72 languages.
AI Search: A search engine that can search, ask questions, and perform tasks.

Chen Hang said, "DingTalk in the AI era should serve real-world work scenarios." This was one of his most public appearances. However, the night before the press conference, he was photographed by netizens patrolling the DingTalk campus in the early hours of the morning, which made him a trending topic on social media as the "most eccentric CEO."

On the day of the press conference, DingTalk also disclosed a set of key data:

The number of enterprise organizations on DingTalk exceeds 26 million, of which over 190,000 are paying users. 79% of the 5,191 A-share listed companies use DingTalk. The number of AI applications on DingTalk has reached 1.41 million.

DingTalk's customer satisfaction rate, which was initially 30% after Chen Hang launched the "Going Down to Earth" campaign in April, has been improved to 80% after he reorganized the customer service team and established three new core teams for data engineering, model training, and performance evaluation. Costs have decreased by 90% simultaneously.

Besides these data, the most noteworthy is DingTalk ONE.

DingTalk ONE was positioned as "DingTalk's new entry point in the AI era" and was considered the core of AI DingTalk 1.0. At the time, DingTalk ONE was the flagship product that was expected to deliver great success.

This is a project that Chen Hang started developing in April. From April to August, DingTalk ONE went from project initiation to launch in less than half a year.

However, this highlight did not last long.

4. ONE's Midsummer and Late Autumn

ONE is not a product that is slowly refined.

The article "Inside the Nail" includes a recap of ONE's lifecycle: ONE's lifecycle began to take shape in April 2025, was first publicly revealed at the launch event on August 25, and its peak DAU stabilized at around 3 million. It was the first project Chen Hang promoted after his return, touted as "AI-native".

However, ONE is also a project with extremely high turnover. Yousu's design leader left in the second week, and the senior colleague who contacted and recommended her to the team in the fourth week was also transferred to another department. Only three product managers at ONE stayed for more than three months, and Yousu was one of them.

ONE has a strong design DNA, with the first head of the design center being the person in charge. However, as the project moved into its operational phase, the initial "card" format gradually evolved into "displaying all important content on one screen".

This was clearly a failed attempt.

After DAU reached 3 million, the retention rate plummeted.

In early 2026, ONE was split up.

It is worth considering that the rise and fall of ONE is not a simple story of product failure.

Its "summer" corresponds to the urgency of Alibaba's AI toB strategy—Alibaba needs a new entry point to carry on the narrative of the Agent era. But its "late autumn" also stems from this; from its inception, it was destined to meet the dual standards set by the group: it had to quickly release products to demonstrate its determination in the AI strategy, while also accepting the physical laws of retention rates for AI-native products.

When these two curves intersect, the product always loses.

There's a line in "Being Inside DingTalk" that aptly describes the cost of this efficiency: it's eager to create a new entry point, eager to prove that DingTalk isn't old.

DingTalk is not old. With 800 million users, 26 million enterprises, 1.41 million AI applications, and 190,000 paying organizations, it remains the number one in China's office collaboration sector.

But Chen Hang doesn't want to be number one; he wants to be "number one in the AI era."

5. The day the nail was broken

March 16, 2026.

Wu Yongming sent an internal letter to all employees announcing the establishment of the Alibaba Token Hub (ATH) business group. This business group has the core objectives of "creating tokens, distributing tokens, and applying tokens," and is personally led by Wu Yongming.

It integrates five major business units: Tongyi Lab (Qwen Big Model), MaaS Business Line (Bailian), Qianwen Business Unit (C-end AI Assistant), Wukong Business Unit (B-end AI Native Working Platform), and AI Innovation Business Unit.

Among them, the "Wukong Business Unit" is making its debut.

24 hours later. On the morning of March 17th, the AI DingTalk 2.0 annual new product launch conference was held.

On this day, Chen Hang launched "Wukong," the world's first enterprise-level AI-native work platform. Its promotional image features a cartoon version of Sun Wukong holding a staff, standing in the center of a group of shrimp soldiers, a scene full of metaphor.

On stage, Chen Hang said, "Today, we've broken down DingTalk, rebuilt it with AI, and created 'Wukong' (the Monkey King). In the past, people used DingTalk to work; in the future, AI will use DingTalk to work."

"Break" is not a figurative language.

DingTalk rewrote its underlying code and fully implemented CLI (Command-line Interface) so that all capabilities can be invoked and operated by AI, rather than being simulated clicks.

"Communication is execution"—in a DingTalk group, a simple sentence like "Generate last week's sales report and synchronize it with management" can trigger Wukong to automatically pull approval workflows, attendance, and CRM data, generate a report, and push it out, all without requiring any manual interface operation.

Wukong simultaneously released ten industry solutions for OPT (One Person Team). This is the world's first solution to transform AI Skills from a technical concept into industry-level, out-of-the-box products, covering ten scenarios including e-commerce, cross-border e-commerce, design, development, and retail.

One detail worth noting is that Alibaba named its AI DingTalk 1.0, released on August 25th last year, "Fern," its 1.1 version, released on December 23rd, "Magnolia," and its 2.0 version, "Monkey King."

"Fern" represents breaking through the soil. "Magnolia" represents new life. And "Sun Wukong" represents being refined.

These three version names essentially represent Alibaba's three-stage assessment of the maturity of AI-powered DingTalk. From "breaking ground" to "development," DingTalk took less than seven months.

In this way, "Wukong" became a new entry point for Alibaba's AI toB strategy.

It's worth noting that the generational shifts in Alibaba's B2B entry points have never been smooth transitions. Each iteration of the entry point is accompanied by the repositioning or marginalization of the original business.

In 2025, the primary consumer-facing (C-end) entry point was Quark. In 2026, the primary C-end entry point was Qianwen App. In 2026, the primary business-facing (B-end) entry point was Wukong. DingTalk was no longer touted as the "number one B-end entry point," but instead became a platform for Wukong.

Chen Hang's role changed from "the person who reshaped DingTalk" to "the person who shattered DingTalk".

This is a highly dramatic transformation within the organization.

6. These 7 days in June

On June 4th of this year, the viral article "Being Inside DingTalk" appeared on Alibaba's internal network.

This is not just an ordinary emotional rant. It is a well-structured and logically rigorous organizational critique report. Using the entire process of the strategic AI project ONE, from its inception to its hasty shutdown, the article systematically exposes deep-seated problems within DingTalk, such as vicious internal competition, autocratic decision-making, meaningless overtime, and mechanical performance evaluations.

Then, on June 8th, a second long article appeared. Ma Ruila (Wang Jiamin), former vice president of DingTalk, published "Being Outside DingTalk" on her personal WeChat official account.

In his article, he said, "I sincerely hope that Wuzhao can lead DingTalk to regain its former glory, but the price should not be that everyone sacrifices their working hours to burn out. In this era, hard work and diligence are certainly important, but flashes of creativity are equally important."

These two articles, one internal and one external, completely exposed DingTalk's internal problems and became a key trigger for the Alibaba Partners Committee's post.

Then on June 10th, the Alibaba Partners Committee posted an article on the company's intranet titled "Company Culture is about Compassion, Loyalty, and Growth." The post harshly criticized the management style of the DingTalk team, directly stating that this "is not what Alibaba culture should be like."

"Under no circumstances, no matter how urgent the task, the management style of the DingTalk team mentioned in the post should not exist. This style has never been the direction advocated by Alibaba's culture, nor is it what Alibaba's culture should be like," the Alibaba Partners Committee wrote in the article.

This is a rare instance in Alibaba's 27-year history where partners have publicly spoken out about internal management issues within a single business line.

Twenty-four hours after the post was published, on June 11, Alibaba announced a management reshuffle for DingTalk: Chen Hang stepped down as CEO, and Chen Yusen, a tech geek born in 1992, took over.

It is worth noting that when Chen Hang was an intern at Alibaba in 1999, Chen Yusen, born in 1992, was only 7 years old.

7. The successor in 1992

Chen Yusen was born in 1992.

A tech geek who rose to fame at a young age. He has won numerous top-tier computer competitions both domestically and internationally. At the age of 22, his cybersecurity company, Changting Technology, was acquired by Alibaba Cloud.

In 2025, he started his own business within Alibaba Cloud, leading the development of the AI Agent product MuleRun.

Born in 1992, a Forbes Asia “30 Under 30”, acquired by Alibaba Cloud, founder of MuleRun – these labels outline a successor with a style completely different from Chen Hang.

Chen Hang's labels are "Dean of the Asylum", "Wu Zhao", and "BE CRAZY".

Chen Yusen is labeled a "tech geek." Although his management style is still evolving, the products incubated by the ATH business group at the same time have already given a signal: Happy Horse, Happy Oyster, MuleRun, Qoder—these names are new footnotes to Alibaba's organizational paradigm of "small teams, young talent, and respect for individuals."

In a post the previous day, the Alibaba Partners Committee emphasized: "People are Alibaba's most valuable asset, and cultivating and inspiring people is the responsibility of every leader. In the AI era, when machines can replace many things that many people can do, people become our most valuable asset. In the AI era, we need to uphold our values of compassion and integrity, and we need to jointly cultivate an open, inclusive, and diverse work culture."

DingTalk's true position within Alibaba's AI toB strategy will not change due to Chen Hang's departure. The strategic importance of the Wukong business unit will also remain unchanged. It remains one of the five core segments of the ATH business group, serving as the application link in Alibaba's entire "Token creation, Token delivery, and Token application" chain.

Alibaba's complete AI toB matrix—Qwen (basic model) + Alibaba Cloud (computing infrastructure) + Wukong Business Unit (B-end application entry point) + Pingtouge (self-developed GPU)—has taken shape.

Chen Hang's 437 days of hard work paid off: the "Agent OS/CLI-based/Wukong Platform" he left behind is the true foundation of Alibaba's B-end A-entry point.

But what he paid for this upgrade was not only the technical and commercial costs, but also the organizational, personnel, and cultural costs.

Clearly, DingTalk's script will be rewritten by someone else.

The entrepreneurial spirit of Lakeside Garden has never been the private property of any one person. It will return repeatedly under different names, with different products, and in different organizations—as long as the organization is willing to leave a door open for it.

In 2014, Chen Hang and five others set off from Lakeside Garden.

In 2026, DingTalk will need to find its own lakeside garden again.

Xiaomi releases and open-sources its exploratory AI programming assistant, MiMo Code.

Popular Articles