Dialogue with PANG Dawei: A small team of ten sails into the blue ocean of "data agents" | NEXTA Innovation Night Talk

Format adjustment, design layout, complex functions... A large amount of time is wasted in the "struggle" with tools, rather than focusing on the content itself. This technological threshold has formed an invisible barrier in traditional organizations: those who are proficient in using tools hold the "privilege" of office work, while novices can only keep querying and learning.

Today, AI and large models are impacting this "solidified soil" with unprecedented power.

Under the theme of office efficiency, the "ambition" of intelligent agents is no longer just to be a Copilot, but to become a real digital partner. The practical need is that new - quality productivity needs to be introduced, but these traditional office suites and common file formats cannot be abandoned in the short term.

It is in this vast area of the old - to - new transition that an AI application called ChatExcel has entered the public eye with its unique entry point - handling complex Excel data through natural conversations.

Pang Dawéi, the founder of ChatExcel, is a serial entrepreneur who has been deeply involved in the SaaS industry for ten years and has written Java code for the same period. When the AI wave began in 2023, Pang Dawéi keenly captured the new opportunities brought by the shift of technological paradigms. Two years ago, starting from the Peking University laboratory, he led a young team of less than ten people, mainly composed of post - 1995s and post - 2000s, to develop the first version of ChatExcel.

Although the concept of Agent was just in its infancy at that time, the real demand for AI - enabled office work made Pang Dawéi see the trend of AI - enabled productivity tools, which is also the only way to explore the boundaries of AI capabilities.

The product was launched in February 2023. Within just half a month, it gained a group of loyal users and quickly entered the ecosystems of large enterprises such as Huawei, Lenovo, and Alibaba Cloud. Thus, Pang Dawéi and his team became the first group of people in the domestic market to successfully verify the market demand for practical AI tools.

What makes the entrepreneurial story of the ChatExcel team special is that it doesn't come from a sudden concept, nor does it aim to create an "all - powerful AI". Instead, it starts from an existing pain point in daily office work and promotes a soft landing of people towards AI applications during the efficiency transformation period driven by the technological explosion.

Its inspiration and motivation are also very "plain", originating from an idea of a doctoral partner in the team when he tried to solve the form - processing problems for his girlfriend who is a primary school teacher. Driven by "love", ChatExcel also confirms the common starting point of many star - level entrepreneurial projects: the real needs from relatives and friends around can spark the fire that makes technology accessible to the general public.

Even in the "office trio", Pang Dawéi and his team didn't choose to pursue the more popular PPT - generation track. Instead, they targeted the technically in - depth field of data processing and positioned the product as a "data intelligent agent" from the very beginning, rather than a simple Excel plugin.

In the highly competitive AI application arena surrounded by giants, ChatExcel has found a "breakout" strategy: with a high density of talents, extreme focus, and rapid response, the small team can find a niche in the segmented field and participate in global competition with an open mind.

Pang Dawéi's entrepreneurial experience combines the purity of a technologist, the pragmatism of an entrepreneur, and the forward - looking thinking about the organizational form and human - machine collaboration mode in the AI era.

In this NEXTA Innovation Night Talk, we had a conversation with Pang Dawéi to jointly explore how an AI application can grow from real needs and how a small team can find its own certainty in the new wave.

Here is the transcript of the conversation, sorted out by 36Kr -

I. From SaaS to Agent: A Turn That Is "More Fun but Also More Stressful"

NEXTA Innovation Night Talk: Why did you decide to shift from the SaaS industry to the AI track when the AI wave came?

Pang Dawéi: I have a technical background and have written Java code for ten years, so I have a natural sensitivity to the emergence and application value of various new technologies. When I was engaged in digital marketing entrepreneurship before, I experienced the new opportunities brought by the changes in mobile traffic from Weibo to WeChat and Douyin. So when the large - model technology became relatively mature, I also realized that AI would surely bring new opportunities. In addition, the SaaS industry has encountered certain growth bottlenecks in the domestic market, while AI just provides a new growth direction.

NEXTA Innovation Night Talk: How would you describe this AI - related entrepreneurship compared with your previous serial entrepreneurial experiences?

Pang Dawéi: There are two keywords for this entrepreneurship. The first is more fun, and the second is more stress. It's fun because ChatExcel, which we developed, is an efficiency tool used by many C - end users, and we have received a lot of positive feedback from them. Compared with my previous To - B business, this gives me a greater sense of achievement. The stress comes from the higher technical threshold of AI entrepreneurship. The team needs a higher density of technology. For example, there are several Peking University doctoral students in our team, which is completely different from the technical barriers required for previous SaaS entrepreneurship. Entrepreneurship in the AI era is completely different from the past, with significant differences in customer groups, team composition, and technical routes.

NEXTA Innovation Night Talk: Why did you choose "Excel + AI" as the entry point for entrepreneurship?

Pang Dawéi: First of all, ChatExcel was born in the Peking University laboratory. Two of my partners had been researching this technical direction during their master's studies. Moreover, we noticed a common need. Everyone around us was complaining that Excel was difficult to use, and the complex function operations in Excel were a common pain point for users. Our team already had a long - term technical accumulation, so we combined it with large - model technology to launch ChatExcel. There is also an interesting background. One of our partners promoted the early development of the product to help his then - girlfriend solve the actual problems in form processing. So, to some extent, three factors contributed to the birth of ChatExcel: technical accumulation, real needs, and love.

NEXTA Innovation Night Talk: ChatExcel is positioned as a "data intelligent agent". How do you understand this positioning?

Pang Dawéi: From the first day of entrepreneurship, we believed that our expectation for ChatExcel was not just an Excel plugin. We don't focus on format preservation but on the processing of file data itself. At the same time, our technical framework is designed in the way of an intelligent agent, which is also different from the traditional plugin concept.

If we just wanted to develop a plugin, we wouldn't need to do it in the future because WPS, DingTalk, and Feishu are all building their own collaborative document, office application, and plugin systems, which are the "territories of large enterprises". We think about this from the dimension of data models. In the future, data - processing methods will become more and more intelligent. ChatExcel is developed based on this concept. The current tool form that makes Excel more intelligent is just a start to adapt to the current data - processing habits of ordinary people.

NEXTA Innovation Night Talk: How did you consider the commercialization path of starting from C - end users and gradually transitioning to B - end users?

Pang Dawéi: We decided to start from the C - end for two main reasons. First, our product comes from the actual needs of C - end users and is naturally designed for them. Second, based on my previous experience in both C - end and B - end entrepreneurship, I judged that the commercialization path from the C - end to the B - end might be more mature.

Considering the time point in early 2023, the B - end market had a wait - and - see attitude and strict requirements regarding the maturity, stability, and compliance of AI products, which had an obvious gap with the capabilities of the technology itself and was not sufficient to support the To - B commercialization route of an AI application. Now, it's the second half of 2025, and the soil for AI implementation in the B - end market has gradually matured. Meanwhile, industry milestones and key moments such as the explosion of DeepSeek R1 at the beginning of this year have deeply educated the market. Now, customers no longer repeatedly ask us "What can AI do?" or "How should AI be used?" Instead, they directly ask "What specific problems can you help me solve?"

II. Survival Rules for Small Teams: Focus, Speed, and Niche

NEXTA Innovation Night Talk: How do you view the current competition in the AI productivity tool track? How can small teams find a survival space surrounded by large enterprises?

Pang Dawéi: In the next two years, we will still see a large number of professional - level and lightweight AI productivity tools emerging, and the competition will be very fierce. However, I think every entrepreneurial team should clearly recognize its own advantages. The advantage of our team lies in focus and rapid response. Our resources are relatively limited, so we must focus on core needs and have a faster execution ability than large enterprises. In addition, we are also cooperating with large enterprises in an open - ended way, so entrepreneurial products can also find their own niches in the application ecosystems of large enterprises.

In our view, large enterprises cannot cover all fields and all user groups. As a third - party, we can provide more flexible services. Entrepreneurship is not something to be afraid of competition. Regarding the question of "Will large enterprises develop competing products with ours?", we always assume that they will, but they may not necessarily do well in every vertical model. According to our observation, in the office field, most intelligent agent teams of large enterprises have invested their resources in the PPT track, which may also confirm to some extent that the direction we chose has natural barriers.

Actually, since we decided to start the business, we have been thinking about competition on a global scale. We not only look at large enterprises but also pay attention to what is happening in the United States and Silicon Valley. When we expand our perspective, it doesn't really matter what large enterprises are doing. We believe that as long as our product and technology are excellent, we will naturally get opportunities.

NEXTA Innovation Night Talk: With a team of less than ten people, you quickly entered the ecosystems of large enterprises such as Huawei and Lenovo. How did you achieve this?

Pang Dawéi: The new technological change and application implementation period will inevitably bring opportunities for new entrepreneurial teams to enter the ecosystems of large enterprises. For small teams, the scarcity of the time point is very important. If we start next year, even if our team grows to 100 people, we may not be able to enter. The qualifications and personnel scale of the team will be more strictly reviewed. At present, there are not so many "qualification - related" problems, and it's basically a competition of product and technology. Technological leadership will make these ecosystems need you. Although you still need to seize the opportunity when it comes, at least by seizing the time window, you can get more opportunities.

NEXTA Innovation Night Talk: In an environment where speed and product quality are crucial, how do you manage this very young team?

Pang Dawéi: I think this is an era of youth. I have a personal experience every day when I interview candidates. Next week, an employee born in 2004 will join our team. In this era, older people tend to see more problems, while young people see opportunities. The driving force of the team lies in whether they like what they are doing and whether they recognize the future of AI - enabled productivity. Our organization is very flat, and the secret of management is to conform to human emotions and logic as much as possible. When everyone wants to do this, you don't need to waste too much time on management.

On the other hand, in the AI era, especially for an entrepreneurial team, it's not feasible to simply tell team members what to do. For example, if Qianwen releases a new version tonight, theoretically, our technical team should autonomously evaluate the impact of the new model on our product, whether it covers our capabilities or whether we can iterate new functions based on the new version. This is a completely proactive behavior because when your new product version is launched, users will immediately start using it, bringing a sense of achievement to everyone. This will form a positive feedback system that promotes the growth and exploration of individuals and the team.

NEXTA Innovation Night Talk: With limited resources, how does the team screen and allocate resources?

Pang Dawéi: Our logic is straightforward: we directly face commercialization. Only real needs can make users pay. Needs that users are not willing to pay for or register for are false needs. One interesting thing about product development in the AI era is that users will put forward their own needs. They will give feedback to developers, saying "Product XX has function XX. Why don't you have it? If you can develop it, I'll pay." This is a situation that rarely occurred in previous C - end product development.

There is another dimension for screening needs: whether it is suitable for our model to do at the current stage. Because we will definitely follow the capability boundaries of the model. This is also different from the resource - allocation logic of product managers in the previous era. It's no longer the era of "do it as long as there is a real need". Now we use multiple dimensions to judge: whether the model can do it, whether someone is willing to pay, and what the promotion rate will be. There are too many needs that can be addressed, but we can't address the needs that the current - stage model performs poorly on or that no one uses, as this will lead to a waste of resources.

III. Reconstruction of Future Organizations: The Value of Human - Machine Collaboration Lies in "Newly Added" Value

NEXTA Innovation Night Talk: What specific changes has the emergence of AI tools brought to organizations and work methods?

Pang Dawéi: From a practical perspective, I don't think the complete reconstruction of organizations has fully occurred yet. The most direct change at present is actually the change in the state and mindset of the team: people have started using AI and are constantly finding suitable scenarios to solve problems. This was not the case half a year ago, but it is clearly happening now.

Actually, the most fundamental part of all organizational reconstructions is the reconstruction of the spiritual outlook. When an organization is willing to try new products, "progress" and "innovation" have already emerged, at least bringing new inspiration. At the same time, the internal communication and collaboration methods of the team will also change. AI is like a nail that has been driven into the "wall" of the enterprise. From a single point to a line and then to a surface, it will gradually drive changes in the entire chain. What the final organizational form will be is still unknown, just like people during the Industrial Revolution didn't know what the future organizations would be like. This is a long - term change.

NEXTA Innovation Night Talk: According to your observation, which types of organizations are more willing to accept AI tools and promote change?

Pang Dawéi: Based on my current observation, the larger the organization, the more likely it is to embrace AI. This is easy to understand. People always talk about cost - reduction and efficiency - improvement through AI. What cost can a one - person company reduce? There isn't much. However, for large companies and groups, a 0.01% increase in efficiency can have a huge impact when multiplied by 100,000 people. How much cost can be saved by speeding up the business by one minute? So, the cost - reduction and efficiency - improvement brought by AI governance have the greatest impact on large organizations. Moreover, at present, the application of AI in many routine business scenarios within large organizations is already quite mature. The larger the team, the greater the impact of AI, and at the current stage, the effects of cost - reduction and efficiency - improvement can be directly seen.

NEXTA Innovation Night Talk: What role should AI play in human - machine collaboration in the future? Where are the boundaries of work?

Pang Dawéi: When discussing the boundaries of AI, it depends on whether your application logic is about replacement (Replace) or addition (New). The most important opportunity for generative AI in the past two years has actually been in "addition". Don't think like some previous SaaS products, aiming to completely replace something. Thinking like this is like "pulling out a nail". If the original performance was at 60 points, can it definitely reach 120 points after replacement? In fact, such a logic hardly exists now.

What AI has done well today is mostly things that traditional technologies couldn't do before. For example, text - to - video generation was impossible no matter how much effort was put in before, but it can be easily achieved today. So, in my opinion, the future opportunities for AI still lie in doing new things, creating new needs, and taking over the needs that were previously abandoned due to inability, rather than replacement. This is also the meaning of human - machine collaboration. If something that a ten - person team couldn't achieve in the past can now be accomplished by adding an Agent to an employee