HCLTech modernizes COBOL to Java with AI + CAST. Boosts accuracy 90%+.

CAST makes large-scale COBOL-to-Java modernization more accurate and efficient for HCLTech by equipping its AI platform with deterministic, application-wide context.

Challenge

Organizations with large COBOL estates face mounting modernization pressure: rising mainframe licensing and infrastructure costs, a shrinking pool of mainframe talent and the complexity of interdependent applications spanning millions of lines. Traditional programs run for years and often fail at scale.

AI and LLMs promise acceleration, yet accuracy degrades as system complexity grows – leaving enterprises without a reliable, realistic path off the mainframe.

Solution

HCLTech embedded CAST Imaging’s deterministic context into its iLIT-AI discovery and analysis platform via the MCP server. In a March 2026 pilot on a nine-million-line mainframe application, the combined approach extracted functional specifications – call flows and business rules – then generated idiomatic Java rather than “Jobol.”

CAST Imaging’s system-wide map grounded the AI, while structural analysis surfaced dead code and scoped testing to the endpoints affected.

Results

The pilot confirmed that deterministic, application-wide context measurably improves AI-powered transformation.

Coupling CAST Imaging to iLIT-AI via the MCP server added 23% efficiency and 15% accuracy in analysis (reverse-engineering step), and 20% efficiency in code generation.

Across both steps, the approach delivered an overall accuracy increment exceeding 90%, with up to 25% dead code flagged for removal and ISO 5055 verification of generated Java.