龙8

In collaboration with Deloitte China and the China Academy of Information and Communications Technology, Digital China, jointly released the blue paper "AI for Process： Enterprise-level Intelligent Process Transformation" at the 2025 World Artificial Intelligence Conference (WAIC). Li Chenlong, the Chief Information Officer of Digital China, systematically expounded the concepts, paths and practices of the blue paper, providing a comprehensive methodology and tool support for enterprises to enter the era of process intelligence driven by AI.

The New Paradigm of Process Reengineering in the AI Era

In the speech, Li Chenlong pointed out that the digital transformation of traditional enterprises was centered around the construction of information systems. For example, CRM, ERP, and HR systems, had significantly improved business standardization and operational efficiency. However, as the number of systems increased, enterprise processes were fragmented among different systems, forming data silos, which made it difficult to meet the present demands for flexible response and intelligent decision-making. Therefore, AI is urgently needed for intelligent upgrading.

He stated that the current large model technology had surpassed human capabilities in multiple professional fields, but the implementation of AI in enterprise scenarios still faced numerous challenges. Although general large models are powerful, they lacked the enterprise-specific data and professional knowledge accumulation. AI applications in enterprise should follow the path of “general-specialized integration”, using open-source large models and internal enterprise data to jointly build specialized intelligent agents, to achieve a low-cost and highly adaptable AI application system in the future.

The key to building an integrated model that combines general and specialized capabilities lies in grasping the intersection point of business model, technological paradigm and management method - the process. "At present, AI was still at the stage of single-point application and had not yet achieved the reconfiguration and reengineering at the process level. However, “AI for Process” aimed to achieve this goal and promoted enterprises to achieve comprehensive perception, rapid decision-making and continuous optimization."

From concept to implementation: Dual-path driving for process intelligence

In order to help enterprises systematically promote the implementation of AI for Process, Digital China had proposed the "Twin-Drive Model" based on a large number of practices, that was, a promotion method combining "top-down" and "bottom-up".

In the "top-down" approach, enterprises start from strategic objectives and decompose their business processes into a multi-level system from L1 to L5, further identifing AI application scenarios at the L5-level. This approach could systematically optimize the process, but it required high-level drive and overall planning. It could ensure comprehensive coverage of AI application scenarios, avoid scenario omissions or execution breakpoints, as well as ensuring organic connections among various scenarios, and to build a complete AI process ecosystem. While the "bottom-up" approach was closer to front-line practice: it was suitable for local innovation and rapid pilot projects by starting from specific operations to sort out the process chain, find AI entry points in existing processes and gradually accumulate value.

No matter which path was adopted, the final goal was to achieve precise description of the AI scenarios. Therefore, Digital China had proposed the “AI Gene Model” as a unified language to help enterprises precisely describe the roles, input and output, operation rules, and data requirements of each AI scenario. It could ensure the deep integration and sustainable optimization of AI and processes.

During the presentation, Li Chenlong thoroughly explained the application methods and effects of the DT model through the company’s own practical experience and customer cases in the healthcare industry: By adopting a top-down approach, the process of LTC was restructured, and multiple AI agents were constructed and rapidly deployed. The iteration cycle of the intelligent agents was shortened from 90 days to 15 days, and it achieved cross-module collaboration and accelerated business response. Moreover, through a bottom-up approach, the "hidden operational chains" in the process of collecting regulations were identified in a pharmaceutical company. After deploying the agents, each person saved 2.2 person-days per month, which significantly improved process efficiency and human performance.

Technological paradigm shifting: Building AI-native enterprise architecture

Apart from the business model and management methods, the intelligentization of enterprise processes also required the support of a corresponding technological paradigm. Li Chenlong emphasized that current enterprise processes carried out in multiple heterogeneous systems were similar to a blind person touched an elephant, making it difficult to achieve an overall perception.

Therefore, Digital China proposed an AI-native enterprise digitalization reference architecture to help enterprises gain a comprehensive view of the “entire elephant”. Its core consisted of two major platforms: the Agent Middle Platform and the Intelligent Process Workbench.

The Agent middle platform integrated system, data, knowledge and model resources from both inside and outside the enterprise. Through unified scheduling and management, it solved the complexity of intelligent agent production. The intelligent process workbench implemented L1-L5 process configuration, achieved consistency in process execution and system operation, and realized flexible reconfiguration and data accumulation for process changes.

He said, “The Agent middle platform and the intelligent process workbench, as the two key tools of AI for Process, have played a crucial role in the implementation for the vast majority of clients. It enabled the AI capabilities to continuously and rapidly iterate within enterprise processes and sustainably deliver value.”

Looking to the future:
Building AI-ready enterprises, Moving towards Intelligent Symbiosis

The promotion of AI for Processes is not merely a technical reconfiguration, but rather a profound organizational and cognitive transformation. Li Chenlong stated that enterprises also needed to possess "AI readiness capabilities", including the establishment of knowledge governance systems, the reconfiguration of organizational processes, the cultivation of multi-skilled talents, and the collaboration of the industrial ecosystem to achieve true intelligent development.

As an indicator to measure the extent to which enterprises would apply AI in the future, this blue paper had proposed the concept of "AI penetration rate" for the first time-that was, the proportion of AI operations in the overall operations (including both AI operations and human operations) in the enterprise business processes. He mentioned that in the future, the one with a higher penetration rate would develop faster. However, from the overall timeline prediction: AI would still be auxiliary after three years, with a penetration rate of 10%-20%; five years later, humans would command multiple AI to collaborate, with a penetration rate of 30%-50%; ten years later, it would be humans supervising the automatic operation of AI, and the penetration rate might reach 80%.

He concluded: "This is the best of times, and yet it is also the worst of times. How to coexist with AI is a problem that every enterprise must confront."

In the future, Digital China will continue to collaborate with various parties to promote the AI for Process concept to be implemented in more scenarios and industries, which can help global enterprises achieve intelligent advancement in the intelligent competition.