China's software industry has surpassed 15 trillion RMB in scale. The rapid development of AI technology is bringing transformative challenges, with traditional models of R&D, business operations, and underlying technical systems facing multiple bottlenecks. All this has made transformation necessary and urgent.

Data from the Ministry of Industry and Information Technology of China shows that in 2025, China's software business revenue reached 15.48 trillion RMB, with a year-on-year growth of 13.2 percent. Out of this, the revenue of industrial software was 333 billion RMB and the revenue of basic software 214.6 billion RMB.

"The old approach of relying on manpower and project-based delivery is no longer feasible," Jia Zijun, deputy director of the Information Software Institute of the China Center for Information Industry Development, said. The core challenges of traditional software industry include the entire process of development relying heavily on manual coding, and engineers spending much time on drawing sketches and searching historical code.

The project iteration cycle is long and the cost of cross-functional collaboration high. Software enterprises have fallen into the growth trap of "the more they deliver, the higher the labor costs and the thinner the profits."

Gao Tongqing, chairman of the Global Intelligent Internet of Things Consortium, made a judgment on the transformation of business models: "AI agents have significantly reduced the demand for human resources, and the original revenue logic is unsustainable. The industry urgently needs to shift from selling software products to a new business model based on tokens and service results."

"Industrial AI should not only generate text, but also build a foundation model for understanding the physical world to develop industrial AI agents that can autonomously plan and handle undefined tasks," Li Peigen, an academician of the Chinese Academy of Engineering, said. He pointed out that intelligent retrieval tools powered by mechanical engineering foundation models can directly read the dimensions and tolerances of the parts, significantly reducing the manpower required for R&D.

Chen Xin, technical director of Alibaba Qoder, an AI coding platform, revealed that AI agents now generate 60 percent of the platform's code. The product iteration cycle has been shortened from two weeks to two or three days, boosting R&D efficiency by nearly 10 times.

He said in the future, routine business systems will be fully handled by AI agents, while human developers will focus on architecture and top-level innovation, forming a new organizational model of "a small R&D team + a large number of AI digital employees."