On August 25, at Alibaba's Hangzhou headquarters, DingTalk CEO Wuzhao made his first media appearance since returning to the company after a five-year absence. This marked his first public interview since rejoining, coming during his 48th year.

The last interaction was in November 2023, when Wuzhao had left Alibaba to establish Two Hydrogen One Oxygen and began exploring cross-border opportunities.

Wuzhao explained that his deep thinking about AI originated from his entrepreneurial period during an overseas trip. At that time, Two Hydrogen One Oxygen's intelligent pet companion devices were expanding internationally. In 2024, Japanese clients requested Wuzhao to provide "visual recognition" services. When he brought the equipment abroad, he discovered that with existing technology, AI couldn't truly achieve "intelligent recognition."

"At that moment, I realized that the essential approach to product development in the AI era isn't thinking about how AI serves humans, but how humans should serve AI, allowing AI to ultimately work with AI thinking and AI methods," Wuzhao said.

This "revelation" in Japan became the catalyst for Wuzhao to reconsider DingTalk. In the second half of 2024, he continuously engaged with AI industry professionals and CEOs from various sectors, contemplating how to create "DingTalk for the AI era," ultimately forming concrete ideas.

Four months ago, Wuzhao returned to Alibaba and took charge of DingTalk again.

Upon his return, Wuzhao spent half a month "investigating product issues" and "examining organizational problems." On the product side, he found DingTalk too bloated and insufficiently streamlined. He reviewed details across different product lines and required each team to obtain genuine user feedback.

Organizationally, Wuzhao began promoting DingTalk's "innovation transformation," breaking down previous hierarchical structures internally and pushing for flattened management. Any employee, including interns, could directly approach Wuzhao to discuss product ideas and proactively propose "innovation projects." Wuzhao's workstation is like everyone else's - in an open office with a regular computer desk. Those who could convince Wuzhao to approve their proposals could become "innovation group" leaders.

Innovation groups represent Wuzhao's organizational "new initiative" since returning to DingTalk. Currently, more than ten innovation groups have been established within DingTalk, each handling an AI scenario or tool. Notably, some individuals who were not previously Wuzhao's direct reports became innovation group leaders reporting directly to him.

Wuzhao requires all innovation groups to maintain "rapid iteration" and be evaluated according to the "co-creation trilogy." For iteration, DingTalk maintains weekly updates overall, with AI scenarios and tools typically receiving key updates within two weeks.

The co-creation trilogy is a method DingTalk used in its early days: first, having an idea that passes internal validation; second, fully understanding customer needs; third, repeatedly testing with strangers to obtain objective feedback. Only tools, functions, and products validated through this trilogy earn survival opportunities.

"I require them to reject the arrogance of large companies and remain humble," Wuzhao said. When he first returned to DingTalk, he felt the team had limited AI understanding and considerable arrogance (referring to insufficient humility toward clients and users). Wuzhao began taking employees on customer visits. He once brought team members to a furniture brand store at 7 AM for on-site research and led teams to reorganize user feedback and calculate genuine user satisfaction rates. "I think I've become much more gentle. I persuade the team using customer feedback and real data," Wuzhao indicated.

Compared to over ten months ago, Wuzhao has changed significantly. Identity-wise, he has returned to Alibaba and leads DingTalk again; in terms of impression, he has returned to the familiar "charge-leading, straightforward" DingTalk version of himself.

On August 25, DingTalk released Wuzhao's first "report card" after four months back: version 8.0, DingTalk ONE. This is a distinctly AI-powered DingTalk, including AI spreadsheets, AI transcription, and various other functions and tools. Additionally, DingTalk launched smart hardware DingTalk A1 - an AI hardware product that can magnetically attach to phones.

Regarding whether this AI hardware product might face criticism similar to early DingTalk being dismissed as an "attendance tool," Wuzhao believes there are misconceptions about DingTalk: "DingTalk's purpose in attendance management was to make it more humane and convenient. Without DingTalk, wouldn't companies still have attendance systems? They would still use fingerprint attendance." Wuzhao also described DingTalk's future positioning, believing it can redefine new global work methods and that he and his team are attempting to change the real "physical world" through AI.

Regarding his feelings after four months back, how his previous entrepreneurial experience influences today's DingTalk, and how DingTalk maintains team innovation capabilities in the AI era, the following is an edited interview transcript:

Q: For hardware products like DingTalk A1, from project inception to finished product took less than four months total. How was this efficiency ensured?

Wuzhao: During my entrepreneurship with Two Hydrogen One Oxygen, we were also making hardware. DingTalk previously had a hardware center and supply chain partners, so when we decided to make hardware, we had these foundations. We had about 40-50 people invest four months into the DingTalk A1 project, and they probably slept less than 5 hours daily. It was truly entrepreneurial style, entrepreneurial state.

Q: Upon your return, more than ten innovation groups appeared internally. Is the DingTalk A1 project also an innovation group result?

Wuzhao: Yes, they are also one of the innovation groups.

Q: When did you get the inspiration for innovation groups?

Wuzhao: I adopted similar methods at Two Hydrogen One Oxygen and brought this approach back to DingTalk. Essentially, I wanted to strengthen the team's innovation capabilities and eliminate problems common in large companies like arrogance and low innovation capacity.

Q: How do you ensure this innovation capability?

Wuzhao: You must select the right people - understand what kind of people you can attract and what kind you value. You can always sense people with the same qualities as yourself. When I returned, I had all product managers undergo a review round to see how everyone thinks. I carefully observed, listened to them explain their products, then visited all customers. This way, you can identify who has imagination and innovation potential.

Q: When we last spoke in November 2023, did you already have today's ideas about DingTalk's AI transformation?

Wuzhao: I actually figured it out about half a year ago, forming a prototype. The core question I was contemplating was how to approach DingTalk in the AI era. Previously, people's understanding was that AI helps humans solve problems; but my understanding is how humans help AI solve problems. The main subject is AI, not humans - this was a key issue I thought through in the second half of 2024. From DingTalk's perspective, what should DingTalk look like in the AI era? I realized the most crucial aspect is understanding how humans should help AI solve problems, then evolving DingTalk under this logic. Furthermore, I think there's a significant opportunity - Chinese people might be able to define global work methods for the first time.

Q: Regarding your 2024 thinking, how detailed were your thoughts then? Which directions were definite, and which weren't?

Wuzhao: The broad direction was certain, but details were refined during implementation. Some things can't be figured out just by thinking - you only know after doing them. But when I returned, I faced a reality: everyone's AI understanding was limited. Most people's AI cognition remained at the application level, thinking about simple matters rather than how DingTalk should transform as an AI work platform.

Q: Was this due to inertia?

Wuzhao: Inertia might be part of it, and another possibility is that I had been entrepreneurial outside. During entrepreneurship, you gain different understandings about enterprises and platforms. My AI understanding deepened during entrepreneurship. I was thinking about AI universality and platform creation, not just AI applications or functions.

Q: What circumstances in 2024 led you to consider how to approach DingTalk in the AI era?

Wuzhao: I was helping other enterprises with intelligent recognition services during my startup. One product was for "cat intelligent recognition," and overseas clients wanted us to use this device for recognition services. But we discovered technology couldn't achieve true intelligent recognition. I realized then that humans are actually helping AI understand the real physical world - this is a crucial process.

Q: You mentioned that upon returning to DingTalk, you found some team members had limited understanding of AI and products. How did you drive their change?

Wuzhao: I told them not to be arrogant or self-righteous.

Q: When you returned, how many of your former subordinates remained? How did you align understanding?

Wuzhao: This might not be about "former subordinates" but about continuous evolution and growth. For example, when I wanted to upgrade customer service products, their response was about response rates and satisfaction levels, but many answers came from experience rather than real observation and genuine user feedback.

Internet professionals can be quite arrogant sometimes. So I took the team to examine everything line by line, obtain real user feedback, observe real user behavior. We might spend three to four hours on one detail problem. Aligning AI and product understanding can't rely solely on preaching - you must spend time leading the team through practice, and they'll start developing intuition. We have about 1,000+ people to persuade and guide. I think I've become much more gentle this time.

Q: Do your team members debate with you?

Wuzhao: My team members can be quite fierce in debates. I'm absolutely not someone who doesn't let subordinates speak. On the contrary, we encourage everyone, even interns, to fully express their opinions.

Q: Have you ever been convinced?

Wuzhao: Rarely. But it has happened. For instance, we previously had an enterprise project management function. Initially, the team's proposed solution differed from my thinking, and I wasn't optimistic about their approach. But later they brought data and demonstrations, which completely convinced and even shocked me.

Q: I must follow up - you mentioned that innovation groups essentially maintain organizational innovation capabilities. How do you design evaluation methods?

Wuzhao: We return to our original "co-creation trilogy" without needing metrics to prove them. Whether products are useful, customers will tell us. I require my team to observe real details, which requires long-term careful observation. When your product is truly useful, users will actually use it and pay for it. I tell them not to tell me what you think - observe what users think.

Q: How do you ensure they actually observe?

Wuzhao: For example, what time did you actually go to the client's company? If you don't go to customers for research and observation, I can take you there and demonstrate. Then you'll understand. If I take you and you still don't get it, that might mean you're not suitable for this team, which needs people who carefully observe users.

Q: For these innovation groups, what are the key evaluation dimensions for their survival?

Wuzhao: Whether they give their all.

Q: Do you assess this full commitment?

Wuzhao: If they give their all, I can provide backup. Sometimes doing something at the wrong time might not succeed, but that doesn't mean the thing itself is wrong. This spirit of full commitment and tough attitude is precious.

Q: For these innovation groups' specific metrics, is it the user volume of their AI products or tools?

Wuzhao: I think the order is customer satisfaction first, then scale. First, you must satisfy one customer.

Q: What dimensions does the group's senior management use to evaluate you?

Wuzhao: Commercial companies definitely have business-related evaluations, and you have a time period to complete them.

Q: How are these innovation groups specifically established internally? Is it top-down from you or?

Wuzhao: They step forward themselves. The innovation group leaders step forward on their own. They proactively approach me, saying they want to do something. My workstation is together with theirs, so they can come chat anytime. Even if you're an intern, you can come find me.