[Digital Today] LG CNS (CEO Hyun Shin-gyun) has unveiled 'DevOn AIND (DevOn Agentic AI Native Development)', an agentic AI-based development platform optimized for enterprise IT environments.

Recently, 'Vibe Coding', which generates code using natural language, has been gaining attention. However, it has limitations as its functions are focused solely on simple code writing, failing to fully reflect the complex system structures and security regulations of enterprises. In particular, in sectors such as finance, public services, and manufacturing, legacy structures and development standards must be considered simultaneously, making actual on-site application difficult.

To overcome these limitations, AIND adopted an End-to-End method where specialized AI agents collaborate to perform the entire process, from requirements analysis to design, coding, and testing and quality verification. When a user inputs requirements in natural language, agents divided by roles—such as △ requirements analysis and design, △ coding, and △ testing and quality verification—operate sequentially to proceed with development.

For example, when a financial institution adds a new service to its existing core banking system, the analysis and design agent designs the system structure based on the requirements, and the coding agent writes the program according to the financial institution's development standards. Users can focus on reviewing and approving the results, thereby reducing the overall development period.

The core foundation is the 'Knowledge Foundation', which integrates and structures enterprise IT assets. By organizing development standards, security regulations, source code, and various deliverables into an ontology database format for AI comprehension, it supports customized development that has learned the enterprise system and business context.

LG CNS also applied a 'Spec-Driven Development' method here. By performing design, coding, and verification according to predefined criteria, it secured quality consistency and minimized hallucinations and errors.

Legacy modernization features are also included. It can automatically convert old language-based systems such as COBOL into Java, or upgrade Java-based legacy systems to fit modern architectures. It significantly increases productivity by reducing analysis, conversion, and verification tasks that previously took weeks or more down to minutes. Currently, the 'COBOL to Java' feature is being applied to next-generation projects of major domestic financial institutions.

AIND was co-developed with Cline, a U.S. open-source AI coding company. Cline's AI coding agent is recording high growth rates on GitHub, a global development platform, and is spreading rapidly. The two companies plan to expand their business into industries where security and regulations are critical, such as finance, public services, manufacturing, and defense, focusing on the United States, Japan, and Southeast Asia.

An Hyun-jung, Executive Director of Application Architecture at LG CNS, stated, "Based on expert-level AI agents that understand enterprise systems, we will automate the construction and operation of large-scale IT systems to contribute to innovation in customer productivity."