Exclusive Interview with Zhang Tong, CEO of Definesys: AI Coding in the B2B Market Can't Be Done by Small Companies

Text | Su Jianxun

If you were to mark the history of Definesys, a software company that has been established for a decade, the Spring Festival in 2025 would surely be an unforgettable time point.

This was supposed to be a time of celebration, but Zhang Tong, the founder & CEO of Definesys, was deeply worried. He hadn't written code for many years, but using the American AI Coding (AI programming) tool Cursor, he managed to write a web - based CRM system in just two days. After that, Zhang Tong came to the conclusion:

"Tools like ours will definitely be replaced by AI."

Zhang Tong is no stranger to the trends brought about by technological changes. He has worked in the traditional software field for twenty years, dealing with established giants like Oracle and SAP on a daily basis. Around 2020, the low - code boom hit. Software development shifted from the old - fashioned "dirty and tiring" brick - laying style to the Lego - like "drag - and - drop" approach. Definesys also followed the trend, launching a series of low - code products and obtaining financing from institutions such as Baidu Ventures.

However, under the AI Coding paradigm spawned by large models, Zhang Tong sensed a crisis he had never experienced before. "If we don't change, we're just waiting for death," he told Intelligent Emergence straightforwardly.

The reason is surprisingly simple - " (Cursor) has an incredibly user - friendly product interaction mode. All you need to do is state your requirements, then hit enter and click accept, and it will just get everything done for you."

Our conversation with Zhang Tong took place in Definesys' office in the AI Tower on the West Bank of Shanghai. In this building, there is Jumping Stars, which is booming and ranks among the "Six Little Tigers" of large models, and also Lanma Technology, which had layoff and unpaid leave incidents at the beginning of the year. The contrast makes the harsh reality of the AI wave's winnowing process very obvious.

Zhang Tong, founder & CEO of Definesys; Photo provided by Definesys

After the Spring Festival, Zhang Tong urgently called a meeting with the management team. There were also more than a dozen AI product experts from the company at the meeting. Zhang Tong directly threw out a series of questions: "Will AI replace us? Three years from now, how many people will be left in each of your teams?"

"Yes, only about half of the programmers will remain." A harsh consensus was reached at the meeting.

Three months after this discussion, on May 15th, Definesys held a strategic press conference on the 105th floor of the Shanghai Tower, the highest conference hall in Asia. It announced its transformation from an "AI low - code and AI iPaaS provider" to an enterprise - level AI Coding and AI Integration provider, and launched the AI - native enterprise platform - DefineX.Al, including:

DefineCoding.Al (Ruijing): Enterprise - level AI Coding platform

DefineFlow.Al (Ruizhang): Enterprise - level AI Integration platform

Definesys named these two products after animals. Whales are highly intelligent, representing the programming ability of the "Ruijing" AI Coding platform; octopuses have multiple tentacles, indicating the data connection ability of the "Ruizhang" AI Integration platform across systems and multiple platforms within the enterprise.

However, compared with Cursor, which Zhang Tong found "extremely satisfying" to use, Definesys' AI Coding platform "Ruijing" has significant differences in positioning and target audience.

Zhang Tong is used to demonstrating product positioning using four quadrants. On the whiteboard, he drew a horizontal axis. To the left of the axis are 50 million "professional developers", where Cursor, Devin, and Windsurf are targeted; to the right are at least 500 million "citizen developers", and the product form also integrates No/Low Code products. This is where Definesys positions itself.

Zhang Tong divides the AI Coding track into four quadrants; Photo provided by Definesys

"Our company's administrative supervisor majored in human resources, but she configured all the company's reimbursement, personnel, and other systems by herself." Zhang Tong gave this example. He hopes that Definesys' products can enable more business personnel within enterprises to get involved in IT programming through a B2B approach.

In essence, overseas AI Coding products represented by Cursor are still targeted at C - end developers. However, Definesys believes that the IT scenarios of Chinese enterprises are more complex. A single C - end product cannot solve the problems. More integrated AI Coding products are needed, "including Low code and No code tools, complex applications, and programmers are still required for development and debugging."

Similarly, enterprises' demand for connection has long exceeded the single dimension of "system integration". Traditional iPaaS only pursues the integration of internal systems and data within the enterprise. What restricts the intelligent process of enterprises is not only the connection of various internal systems but also the more important task of breaking through the boundary between the enterprise and the external intelligent world.

This is also the reason why Definesys launched the enterprise - level AI Integration platform "Ruizhang" while releasing the enterprise - level AI Coding platform "Ruijing". This product covers core modules such as AI iPaaS, AI large - model gateway, MCP intelligent middle - platform, and Agent development platform. It realizes the intelligent integration of applications and data within the enterprise through AI iPaaS and the connection between the enterprise and the external "world intelligence" through the capabilities of MCP and Agent.

To some extent, Definesys' product positioning is also about consolidating its strengths and competing differently from some players in the market.

AI Coding is becoming a red - ocean for startups. In China, star entrepreneurs such as Su Wen, a former investor from Huachuang Capital, Chen Zhijie, the former head of TikTok's algorithm, and Ming Chaoping, the former core product leader of Yuezhianmian, have all secured large amounts of financing through AI Coding. Overseas, in early May, the AI programming unicorn Windsurf was sold to OpenAI at a valuation of $3 billion.

For Definesys, its ten - year accumulation of enterprise service experience and the resources of over a hundred large - scale group enterprise customers, including BYD, BOE, and Procter & Gamble, are its bargaining chips in the B2B AI Coding track, which coincides with Windsurf's strategy.

If AI Coding represents the "development hub" for enterprises to build intelligent capabilities, then AI Integration is the "bus platform" for enterprises to connect with the world's intelligence. These two products are Definesys' redefinition of the future enterprise technology architecture:

The intelligent boundary of an enterprise should no longer be limited to its own systems but should include real - time interaction and intelligent collaboration with large AI models, Agents, and ecological services. Only when AI Coding and AI Integration form a two - wheeled drive can the closed - loop of enterprise intelligent application interaction be realized.

Previously, Anshul Ramachandran, a founding member of Windsurf, shared his thoughts on enterprise customer acquisition: From the very first day of building the product, the team had in mind targeting large enterprises in traditional industries. This is because compared to competing for individual users in the red - ocean market, enterprises are the main payers for software. Winning a large enterprise customer is equivalent to winning thousands or even tens of thousands of individual users.

In terms of team changes, Zhang Tong also said that the development of AI Coding will reduce the proportion of programmers in the team but will require an increase in personnel who understand sales and strategies. "It means that with the same team size, the number of customers we can serve can increase tenfold."

During the interview, when talking about the strategic meeting at Definesys after the Spring Festival, when the artificial intelligence experts present reached the conclusion that "at least half of the team will be replaced by AI", I asked Zhang Tong if everyone would be very disappointed when reaching such a consensus.

"Definitely not." Zhang Tong replied firmly. "On the contrary, everyone is excited because a new world is coming."

The following is the core content of the interview between Intelligent Emergence and Zhang Tong, the founder & CEO of Definesys, edited and published:

1. "If we still rely on a large number of programmers to develop software three years from now, the company will definitely fail."

Intelligent Emergence: At the press conference, I heard you talk about a transformation, "from a software company to an artificial intelligence company". How was this plan formulated? It sounds very urgent.

Zhang Tong: Yes, if we don't change, we're doomed.

Zeng Ming (former Chief Strategy Officer of Alibaba) said that formulating a strategy is about "thinking about the next ten years, looking at the next three years, and taking action in the next year". I started to think, will SaaS and low - code still exist three years from now? I don't think so.

People often ask me, if AI can write programs in the future, what's the point of having you? To be honest, if you had asked me in 2024 whether AI could replace us, I would have said no.

Because the digital construction of enterprises is extremely complex. There are ERP/CRM/MES/WMS (all are IT systems for different aspects within an enterprise) ... and the construction of the entire open platform and the underlying PaaS structure. How can you use AI to replace such a complex environment?

Intelligent Emergence: What made you change your mind?

Zhang Tong: During the Spring Festival, I spent some time researching Cursor. Once I used it, I realized that tools like ours would definitely be replaced by AI.

I hadn't written code for many years, but with Cursor, I wrote a CRM system in two days. I even added a chatbot to it, which can conduct public opinion analysis, retrieve data from Internet platforms, and associate sales leads.

The key is that Cursor's interaction mode is so great. All you need to do is state your requirements, then hit enter and click accept. You don't need to do anything else, and it will just get everything done.

At that time, I thought that any programming applications with interfaces that require drag - and - drop operations would be wiped out by Cursor. It's as if the software world has been reconstructed.

Intelligent Emergence: Is it really that absolute? Many programmers think that the code written by AI still has a lot of problems.

Zhang Tong: The quality is indeed not high. Human programmers usually think in a modular way, abstracting many projects into "building blocks" that can be called when needed. But AI is result - oriented, directly generating the final version. If there are bugs, you have to go back and fix them.

Of course, there are still issues such as system permission management, security testing, and code inheritance that AI can't solve for now. But looking into the long - term future, say three years from now, can these problems be solved? I think they can.

The virtual world can be reconstructed by AI. Eventually, the interface people use will be in the form of AI - based

CUI (Conversational User Interface, completing command - line user interactions through conversations), rather than writing programs manually. If we still rely on a large number of developers to develop software three years from now, the company will definitely fail.

Intelligent Emergence: What do these observations mean?

Zhang Tong: After the Spring Festival, we held another meeting and finally reached the conclusion that we will definitely be replaced by AI, and at most half of the team will remain.

Intelligent Emergence: Will people be very discouraged because of this?

Zhang Tong: No, they will think that a new world is coming, and they are at the forefront of this new world. This is a huge advantage.

2. "It's easy for small companies to go bankrupt if they try to do AI B2B."

Intelligent Emergence: Many people in the industry talk about AI and low - code and think they can coexist, but your view that "AI will replace low - code" seems more radical.

Zhang Tong: Actually, everyone has a sense of crisis, but some people just don't admit it. They're afraid that admitting it means they'll really be disrupted by AI and that customers won't buy their products.

We're the first in this industry to say that AI will definitely replace low - code. If someone says it won't, they're going against the historical trend.

Intelligent Emergence: What are the product positioning and target audience of our products?

Zhang Tong: This has to do with the analysis of the AI Coding track, which can be divided into four quadrants (he drew it on the whiteboard). Well - known products like Devin, Cursor, and Windsurf target professional developers, that is, 50 million programmers. We're in the No/Low Code quadrant, targeting the digital IT departments within enterprises, ITBPs, and even a wider range of business personnel.

For example, our company's administrative supervisor majored in human resources. She configured all the company's reimbursement and approval systems by herself, and the finance department configured the financial system on their own. They haven't learned programming, but they just need to know how to use low - code products.

This also has to do with the IT environment in China. Cursor and Windsurf are indeed good products, but within enterprises, there are often cross - platform interactions between software such as ERP/CRM/MES. Chinese companies also have a large demand for private deployment. This is not something that a single product can solve. It's an engineering problem.

So the product demand in China must be in an integrated AI Coding form, not a single - type one. It requires some Low code and No code tools, but different from the current product forms. The interaction models of these tools will be more intelligent, a combination of CUI and GUI, and there also need to be some programmers to expand and debug these things to handle complex tasks.

Intelligent Emergence: There are many entrepreneurs doing AI Coding recently. How do you view your position in the competition?

Zhang Tong: We have an absolute advantage. First of all, in terms of business model, we'll do SaaS deployment, but we'll definitely also adhere to local deployment because that's what Chinese enterprise customers need.

Many early SaaS providers looked down on us (those doing local deployment), but we have our own judgment on what customers really need and how the market will develop. Now, those SaaS providers are also starting to return to local deployment.

Moreover, in B2B competition, relying solely on AI Coding is not enough to meet the complex business needs of enterprises. AI Integration complements this aspect. Combining AI Coding capabilities with AI Integration capabilities truly enables the closed - loop implementation of enterprise intelligent applications.

If a small team wants to do AI Coding on its own, it might work for the C - end market, but enterprise customers will definitely not buy it. Because all its capabilities are based on individual programming and don't solve problems such as security, permission management, environmental compatibility, and post - maintenance. This is an extremely difficult and long - term accumulation - required track. It's easy for a small company to go bankrupt if it tries to do it.

However, for Definesys, the product transformation is relatively easier. On the one hand, many large enterprises are already using Definesys' products. All we need to do is to better and faster integrate the AI Coding part into the products. On the other hand, since we deal with various problems in enterprise digitization every day, we understand the problems customers face very well. So using AI to solve these problems is what we're better at.

But we also have our own difficulties. That is, how to transform our organization into an AI - native organization and product. This requires a lot of effort in terms of strategy, execution, organization, and culture. In this regard, startups actually have an advantage.

Intelligent Emergence: Internet giants like ByteDance, Tencent, and