Kiro Agent and Harness Engineering: Automating Large-Scale Knowledge-Based Migration at LGU+

LG U+ carried out a large-scale migration targeting about 100 services to transition its complex legacy systems from an on-premises environment to AWS Cloud. This project was not a simple lift-and-shift but a highly challenging effort involving dozens of tasks, including Java version upgrades, DB migration (Altibase/Oracle → Amazon Aurora), and storage transition (S3).

The project is broadly divided into two phases. In the Assessment phase, an AI Agent analyzes the source code to derive conversion items (Action Items) and generate design documents, while in the Migration phase, actual code conversion and verification are performed. In particular, complex tasks requiring business logic judgment are carried out by developers working in collaboration with Kiro, an AI coding assistant.

However, there was a problem where the quality of results varied depending on each developer's AI proficiency and prompt quality. To address this, the Harness Engineering environment was incorporated to reflect the organization's technical standards and architecture conventions in the AI's context, aiming to reduce the capability gap between developers and achieve migration of consistent quality.