AI-Led Modernization Preserves Critical Knowledge

EPAM used its MFLens platform, part of AI/Run™.Tools, to analyze the firm's COBOL programs, JCL, stored procedures and database schemas, mapping business logic, application relationships and data flows across the estate. Instead of asking SMEs to document 994 programs manually, EPAM provided them with AI-generated business process flows, dependency maps, data lineage and technical metadata to review — freeing experts to spend their time confirming logic and adding the context only years of experience could provide. The result was a detailed, traceable picture of business logic and rules that had been hidden inside the code for decades. In four months, the team delivered 312 features, enhancing the tool to the customer specific use cases.

With reverse-engineering outputs in hand, EPAM worked alongside SMEs to validate findings and capture the context that lived beyond the code itself. What had existed only in millions of lines of COBOL and in a handful of specialists' heads became a centralized, searchable knowledge base — business rules, dependencies, lineage and application relationships, documented in a way future teams could actually use. That knowledge base reduced the firm's reliance on undocumented institutional knowledge while creating a durable foundation for future engineering, support and modernization initiatives.

With the current state understood and validated, EPAM used AI-assisted and agentic engineering workflows to translate business rules, dependencies and process knowledge into future-state designs, domain models, API specifications and implementation-ready technical artifacts. In the next three months, the program was architected and designed for multiple features giving delivery teams a current, transient, target-state architecture and a clearer “definition of ready” for implementation.

Modernization planning went beyond technical analysis. EPAM worked with stakeholders to map legacy programs to business capabilities and score them against a framework balancing business value, technical complexity, dependency risk and readiness. That gave leadership an objective basis for sequencing investments instead of guessing — and the framework proved repeatable enough that it was subsequently applied to three additional portfolios, encompassing more than 6.2 million lines of code.