Story by NUPIAO |
Edited by NUPIAO | Wen Shuqi
By 8 p.m. on June 11, most of the office towers in Hangzhou’s “Qincheng Park”—where DingTalk is headquartered—were already dark.
The street-level shops below were still buzzing, with someone taking their dog for a stroll. Only a handful of people trickled out of those boxy buildings. But for DingTalk today, that kind of quiet evening feels completely foreign.
“Our team’s schedule shifted from 9-10-6 (9 a.m. to 10 p.m., six days a week) down to 9-9-5 (9 a.m. to 9 p.m., five days a week).” Lin An (a pseudonym), a DingTalk employee, told NUPIAO recently that his group was clocking off by 7 p.m. with zero late-night meetings.
A massive 75,000-word resignation essay finally cracked open the pressure vessel that had been building up behind DingTalk’s walls.
On June 4, a piece titled “Inside the Nail” started circulating on Alibaba’s internal forum before blowing up across the entire internet. It was penned by You Su (pseudonym), one of the lead product managers for DingTalk’s strategic “ONE” initiative.
The essay laid bare the entire lifecycle of a highly anticipated AI project: from its ambitious launch, surging past 3 million daily active users, to a brutal retention drop-off and eventual rollback. More importantly, it dragged DingTalk’s controversial management culture—characterized by relentless pacing, upward appeasement, and organizational burnout—straight into the daylight.
Readers instantly connected with several painfully relatable details. Take the “Wangshu Operation”: teams were ordered to keep watch on competitors like Feishu. If the lights stayed on across the street, DingTalk staff couldn’t leave. Then there’s the “Daily Delivery” expectation designed to cater to executive whims: “Any request bosses toss into the group chat in the morning has to be boxed up and ready for review by nightfall. Change the UI today? Finalize the copy today? Adjust the rules today? Package it today? Flag the issue today? Pitch the solution today? Honestly, even saying ‘tomorrow’ starts feeling like a luxury.”
Two days later, on June 10, Alibaba’s Partner Committee罕见ly posted a statement on the internal network titled “Loyalty, Compassion, and Growth Define Our Culture.” Using unusually sharp language, the committee condemned DingTalk’s leadership style, stating it “doesn’t align with Alibaba’s core values. No matter the circumstances or how tight the deadlines are, the kind of management practices described in the post simply have no place here.”
The very next day, leadership changes were officially confirmed. Marking a historic first, an entire management overhaul was triggered by a single employee’s critical long-form post.
DingTalk founder Chen Hang, known internally by his alias Wu Zhao, stepped down for the second time after more than a decade guiding the product he originally built.
This departure capped a turbulent 437-day stretch since his return on March 31, 2025.
Competition & The Anxiety Factor
Roughly around 10 a.m. on June 11, Lin An heard the first gasps of surprise ripple through the floor.
Colleagues huddled near adjacent desks, eyes glued to their screens, whispering frantically. Several others stood up abruptly from their chairs.
Once he pulled up the browser himself, the news hit: Wu Zhao had officially resigned as CEO.
An engineer from another Alibaba BU later recalled being on a video call with DingTalk staff when the connection suddenly dropped. Moments after the announcement leaked, the audio feed on the other end erupted into heated debate.
There was no official internal memo. No all-hands blast. Someone tossed a firework GIF into the main DingTalk group chat, only to pull it back seconds later. Others quietly started organizing farewell dinners.
A former employee shared with NUPIAO that ex-DingTalk colleagues rarely bother posting in WeChat groups anymore. Before, using external messaging apps could actually trigger mandatory self-criticism sessions.
Despite the shake-up, DingTalk still commands the largest user base in China’s collaborative office software market.
We’re talking 800 million registered accounts, 26 million enterprise organizations, and roughly 200 to 230 million monthly active users (MAU) leading the pack. But growth has effectively flatlined.
FY2025 (April 2024–March 2025, during former CEO Ye Jun’s leadership), despite shifting focus toward large-client customizations, the capital poured in never materialized into substantial commercial returns. That same year, DingTalk’s subscription revenue sat at roughly 3 billion yuan. In stark contrast, Feishu, whose user base was merely ~15% of DingTalk’s, projected an ARR exceeding $300 million in 2024 — equivalent to about 2.1 billion yuan, or 70% of DingTalk’s matching period revenue.
Chen Hang previously revealed during DingTalk’s 10th-anniversary event that as of August 22, 2025, the platform hosted over 26 million enterprise organizations. Of those, paid accounts topped 190,000—a conversion rate stubbornly stuck below 1%.
Feishu, operating at a smaller scale but moving faster commercially, rapidly closed the revenue gap through higher monetization density. Solving this imbalance was undeniably Chen’s first priority upon returning.
From a product standpoint, Feishu and DingTalk come from completely different worlds. Feishu leans hard into docs and multi-dimensional spreadsheets. One enterprise exec who used both platforms noted that Feishu’s grid system is the killer feature widening the competitive gap. Because every core module is built in-house and deeply integrated, the overall experience feels remarkably consistent and fluid.
DingTalk, on the other hand, has always played the “build infrastructure, foster ecosystem” game. While feature-complete, that breadth introduces friction. Many advanced tools—docs, sheets, approval workflows—are either pulled in from partners or bolted on via acquisitions. Former CEO Ye Jun publicly acknowledged that DingTalk’s search experience degraded over time, proving that stacking features doesn’t automatically translate to better productivity.
Then comes the elephant in the room: the AI era is actively dismantling traditional software models, leaving everyone scrambling.
Gartner’s early 2026 forecast predicts that by year-end, 40% of enterprise apps globally will embed task-specific AI Agents, a massive jump from just 5% in 2025. These agents don’t just assist—they deconstruct tasks and execute deliverables autonomously.
That reality sends shockwaves through every productivity suite: if AI Agents can handle scheduling, meeting notes, information syncing, and even basic approval decisions, the entire value proposition of human-centric collaboration tools gets fundamentally questioned.
The AI arms race has become the central battleground for DingTalk, Feishu, and WeCom. That’s Chen’s second major anxiety point.
In 2025, Alibaba locked in its AI strategy, designating DingTalk as the group’s primary To-B gateway.
Last April, Chen made headlines by predicting the death of conventional software, forecasting a future where apps become “daily disposables”—instantly generated on-demand and discarded afterward. His blueprint for navigating the AI wave? Transform DingTalk into an Agent OS tailored for work. The mission: run and orchestrate AI agents, handing the actual labor to machines while humans step back to focus purely on decision-making.
Chen bets everything on AI velocity. He’s vocal about the mantra “go AI or get left behind,” insisting that only leaders who truly believe in the tech can drive the necessary reforms. Without that conviction, he argued, DingTalk would get crushed by faster-moving rivals.
Enterprise clients definitely felt that urgency. One DingTalk account told NUPIAO that in his first month back, Chen flooded the field with customer success teams running hyper-detailed satisfaction surveys, essentially trying to tear down and rebuild the playbook from scratch. By sheer execution speed, the operation moved lightning fast.
In my view, Chen’s directional call wasn’t necessarily wrong. DingTalk is fundamentally a To-B tool built to solve pain points for organizations and managers. The fact that the “Wukong” platform failed to win over paying customers actually proves he was onto something: it leaned too heavily into end-user experience while ignoring what decision-makers actually need. “Wukong,” launched in March, replaced the “ONE” entry point as DingTalk’s AI-native hub.
Chen joined Alibaba in 1999 as the company’s very first intern, and that early imprint shaped his worldview. He’s frequently cited the platform’s original mandate: “Free the world-class work methods pioneered at Alibaba to China’s 43 million SMEs.”
Insiders close to Alibaba tell NUPIAO that internally, Chen remains a polarizing figure—but far from toxic. “His track record speaks for itself. Stepping down isn’t a firing; it’s a strategic pause to steer clear of the immediate spotlight.”
Departure & Return
To understand what unfolded during those 437 days, we need to rewind further.
Back in January 2015, the first version of DingTalk launched. Features like the “DING!” notification and read receipts nailed exact management pain points, helping it secure rapid adoption among small and mid-sized businesses. It crossed 100 million users in year one and broke 300 million within three years.
Early on, Chen’s “zero-to-one” hustle earned widespread respect inside Alibaba. But let’s be honest—it was forged in what insiders jokingly called the “madhouse.”
After losing the social wars against WeChat with his previous app, Lailai, Chen visited dozens of SMEs in spring 2014. The feedback was clear: he needed to start fresh, building an enterprise workspace from scratch.
The founding crew numbered barely six or seven folks, holed up developing in Lake Garden, earning the “madhouse” nickname from outsiders.
But that “crazy” mentality carried a shadow side: an almost suffocating management style.
Chen’s personality was unmistakable from day one. Working until 11 p.m. nightly, seven days a week, he literally painted “Born to Fight” on the office walls. Team tees flipped that script with “BE CRAZY” printed on the back.
Zi Ning (pseudonym), an early DingTalk contributor, told NUPIAO that Chen’s prior stint living and working in Japan heavily influenced his leadership approach: high pressure, rigid hierarchy, absolute obedience, obsession with granular details, and tight control.
During the scrappy startup phase, that wolf-culture was viewed as necessary survival fuel. But once headcount scaled past thousands, the same playbook started drawing serious pushback.
In 2021, amid strategic splits over cloud-and-Ding integration, Chen left Alibaba for the first time. He subsequently founded HHO in Japan, focusing on cross-border hardware, though progress remained relatively steady.
Four years later, as DingTalk was cemented as Alibaba’s To-B AI flagship, the company bought into HHO’s investor shares to coax Chen back. The recruiter? Wu Yongming, his original internship mentor from 1999.
At the time, outside observers universally labeled it an emotionally sound move. A founder’s homecoming typically injects massive morale into a restless team. And looking back, his return timeline absolutely synced with the industry’s acute AI-era anxiety.
Zi Ning pointed out that Chen’s “second act” actually carried heavier stakes than a fresh appointment. “He might’ve believed his initial playbook was flawless, pushing to prove himself quickly. That urgency often creates blind spots that end up boxing you in.”
Chen responded by accelerating DingTalk’s transformation with intense management force.
According to LatePost, over four months, the DingTalk squad cleared 574 backlog tickets, audited 1,850 historical feature requests, and refined over 20 product lines. NUPIAO learned that internal AI initiatives briefly multiplied into a dozen projects before consolidating into AI transcription, smart sheets, AI search, and the hardware debut DingTalk A1. By execution metrics alone, Chen’s pace was undeniable.
Yet beneath that intensity, certain management habits began distorting.
A departing employee recalled that once Wu Zhao took charge, attendance and discipline standards tightened noticeably. “A lot of expectations lived entirely outside official policy, pushing enough people to their breaking point.”
Inside DingTalk, mandatory camera-on meetings became standard. Deviate and you’d face immediate interrogation. Travel budgets were officially capped at 700 yuan, but actual reimbursements routinely got slashed to 300. Leaving the office past 10 or 11 p.m. drew skepticism. External apps like WeChat were banned outright—only DingTalk communications allowed. That hermetically sealed environment stood in stark contrast to Alibaba’s other business units.
For rank-and-file staff, fading purpose proved equally draining.
One DingTalk engineer posted online about having almost zero runway for long-term builds. Leadership demanded full-stack project delivery within a week. Over six months, he churned through a dozen algorithm pilots, each abandoned after seven days. Either client briefs came in, got built, flopped, and got scrapped—or you pitched upward, only for leadership to pivot direction anyway.
During Chen’s 18-month tenure, two major public launches happened, yet internally, he virtually skipped all-hands town halls to clarify overarching strategy.
Zi Ning painted a clearer picture for NUPIAO: “Department splits and personnel reshuffles rolled out without transparency. Nobody knew where the real strategic north star was. Most reform energy burned into attendance tracking, SOP enforcement, and travel limits instead of concrete directional shifts.”
No one saw a single internal forum thread igniting the whole storm. On June 8, former AI lead Ma Rila published “Outside the Nail,” confirming his May 15 departure and confessing, “I’m losing grip on whether I’m actually shipping products or just burning my body chasing a finish line that keeps moving forward.”
NUPIAO understands that ahead of the public blowup, a steady queue of employees was already processing exit paperwork. Once Chen’s resignation surfaced, those departures temporarily paused.
AI Demands Brains, Not Bolt-Tightening
The incoming boss, Chen Yusen, born in 1992, cuts a distinctly different profile. Pure technologist roots.
He launched cybersecurity firm Changting Tech at 22 before Alibaba Cloud acquired it. Listed in Forbes Asia’s “30 Under 30,” he’s a textbook AI Agent native founder, previously steering development on the MuleRun intelligent agent platform.
Resumé and mindset combined, Chen Yusen forms a complete mirror image to Chen Hang. Taking the reins made him Alibaba’s youngest divisional CEO.
Most staff haven’t worked directly with him yet, but consensus paints him as “undeniably young” and “likely driving stronger innovation momentum.” Beyond the paper credentials, everyone’s really watching whether he can usher DingTalk into a new era equipped with genuinely inventive, long-term AI product logic.
In MuleRun, teams operated lean and aggressively flat. Chen has openly predicted that 60% of current Agent products will vanish by end-2026.
Contrasting Chen Hang and Chen Yusen’s backgrounds highlights a visible shift in Alibaba’s internal playbook: trading rigid legacy structures for younger, creatively agile minds suited to next-gen AI organizational models.
Over the past couple years, Big Tech has gone feral competing for AI talent. On April 16, reports surfaced that Guo Daya accepted a nine-figure total compensation package joining ByteDance’s Seed division, sending tremors through Silicon Alley. Maimai data shows AI scientist/lead average monthly salaries skyrocketed to ¥132,796 between Jan-Apr 2026, leaving peers miles behind.
Beyond top-tier research brains, recruitment pipelines are expanding. ByteDance launched global campus hires targeting frontier tech. Tencent opened rolling summer internships for its 2026 Qingyun program with uncapped comp. Alibaba committed $50 billion toward spring recruiting aimed at overseas grads.
An AI-native startup founder summed it up bluntly to NUPIAO: “The game changed completely. AI runs on cognitive horsepower and ambition, not mechanical bolt-tightening.”
All these signals scream one thing: AI rewards rapid adaptation and genuine creativity. Yet at DingTalk, anxiety manifested through grueling overtime, surveillance-style attendance tracking, and the “daily delivery” sprint cycle. That高压 culture doesn’t equal agility. It actively conflicts with what modern knowledge work requires from humans: aesthetic judgment, curiosity, and creative exploration.
So the real question remains: facing the AI shift, how exactly should a scaled organization actually win?
The board Chen Yusen inherited isn’t exactly warm and fuzzy. DingTalk still dominates China’s collaboration space, spanning 26 million enterprises and ~200 million MAU. Scale aside, structural cracks persist: paid orgs barely crossing 200k, low conversion rates. Rival Feishu leveraged a slimmer footprint to extract significantly higher monetization density.
In March 2026, Alibaba spun up the ATH Business Group led directly by group CEO Wu Yongming. Within it, DingTalk—steered by Chen—launched the enterprise-grade AI-native workspace “Wukong,” positioning the Wukong unit as Alibaba’s primary To-B AI gateway.
With leadership swapping hands, will product priorities pivot? And critically, that high-pressure playbook clearly didn’t spark explosive commercial growth. Instead, it ignited PR backlash. Can the new regime navigate both? Those are the live wire problems sitting on Chen’s desk.
Looking backward, Chen Hang defies simple black-and-white labeling. He delivered undeniable wins, but controversy followed closely.
Strip away the noise and judge pure velocity: the AI transcription initiative burned past 1 billion yuan, partnered with Tongyi Lab for 100 million hours of audio training, temporarily boosted customer satisfaction to 80%, and slashed support costs by 90%. In the AI recorder category, AVC data placed DingTalk alongside iFlytek, PLAUD, Anker, and Newmine in Q1 2026 retail rankings. The performance metrics are real.
Chen’s product instincts ran aggressive. Wukong shipped fast. But let’s be clear: the ONE project’s messy conclusion, plus the heavy-handed culture, form the epicenter of external criticism.
Visibly, DingTalk’s turbulence may cool down now that Chen Hang has stepped aside. The lingering challenge, however, remains unresolved: when AI rewires how work operates, what operating cadence and cultural DNA should a giant like this actually run on? That’s a puzzle demanding serious reflection, not quick fixes.
(Intern Jin Jing contributed to this report)
