Improving Software Quality Through AI-Assisted Discovery in a Large Enterprise System
Large enterprise systems frequently evolve through continuous maintenance activities, changing business requirements, and contributions from multiple teams without corresponding updates to their technical documentation. Over time, this process leads to fragmented knowledge, reduced system understanding, and increasingly reactive maintenance practices. This paper reports the experience of restructuring the mobility module of Paytrack, a large-scale travel and expense management platform. To address recurring operational issues, a dedicated team conducted a systematic discovery and reverse engineering process combining collaborative code analysis, technical story decomposition, abstraction-oriented documentation, artifact generation through artificial intelligence, and continuous knowledge-sharing sessions. One of the main contributions of this initiative was the Abstract Sketch Strategy, an approach focused on representing complex system flows using higher-level abstractions to improve readability and reduce cognitive overload. After four months, the initiative resulted in a 75% reduction in bugs escalated to the development team and a 90.13% reduction in contact rate related to the module. Additionally, the initiative improved collective knowledge sharing and increased the maintainability of the system, making the team shift from a reactive development to a preventive one.
