LG CNS said on the 8th that it released DevOn Agentic AIND (hereafter AIND), an agentic artificial intelligence (AI)-based development platform that performs large-scale IT system buildout and operations.
Recently, "vibe coding," which generates code from natural-language instructions, has drawn attention, but it is seen as limited for use in building large-scale IT systems. Its functionality is confined to code generation, and because it produces code without understanding the structure and context of corporations' systems, it can clash with existing systems or cause partial changes to affect entire codebases. In particular, in environments such as finance, public sector, and manufacturing, it has been difficult to apply because it cannot reflect security rules and development standards.
To solve these issues, LG CNS partnered with U.S.-based global open-source AI coding company Klein to develop AIND. When a user enters requirements in natural language, AIND assigns specialized AI agents—such as a customer requirements analysis and design agent, a coding agent, and a testing and quality verification agent—to collaborate and divide up the entire development process.
For example, when a financial company wants to add a new financial service to its existing core banking system and a user inputs, "Build an automatic transfer service for savings and time deposits linked to the account system," the analysis and design agent analyzes the customer requirements document and designs the system architecture. The coding agent, which takes over the design, generates code in line with the financial company's development standards. An LG CNS official said, "Users can focus on reviewing and approving results, shortening development lead time."
AIND's strength is the "knowledge foundation," which integrates and analyzes the IT information needed for system development. The knowledge foundation is an ontology databases structured so AI can understand corporations' IT information, including development standards, security rules, system source code, and development deliverables. Based on this, AIND learns corporations' systems and operations and leads customized development.
LG CNS applied a "spec-driven development" approach to AIND. By having AI perform design, coding, and verification according to predefined criteria, it can minimize hallucinations (AI illusion phenomena) and errors. The company also said it supports "legacy modernization," which converts systems into architectures optimized for the latest technology environment regardless of programming language, cutting code analysis, conversion, and verification work—from what used to take weeks—down to minutes.
Ahn Hyeon-jeong, executive director in charge of Application Architecture at LG CNS, said, "Based on AI agents with expert-level understanding of corporations' systems, we will automate large-scale IT system buildout and operations to help drive productivity innovation for corporate clients."
LG CNS and Klein plan to expand the application of AIND to IT system buildout and operations projects for global corporations in sectors where security and regulation are critical—such as finance, public sector, manufacturing, and defense—focused on the United States, Japan, and Southeast Asia.