Legacy Software + Agentic Discovery

Legacy software powers the world - from banking to utilities and government. The hardest part is rarely having access to the code. It’s recovering the knowledge around it: what the system really does, the business rules it encodes, the edge cases no one remembers, and the intent that never made it into documentation.

To modernise safely, teams need more than technical understanding - they need functional understanding. A legacy codebase is a crime scene: you have to retrace the steps, gather evidence, and reconstruct the story in plain language that product, engineering, subject-matter experts, and AI agents can all work with.

In this session I’ll share lessons from building Code-to-docs: a software reverse-engineering system that turns existing codebases into living software definitions. I’ll cover the evolution from a fixed LLM pipeline to agentic discovery, the real trade-offs between quality, cost, and speed, and case studies showing how teams de-risked modernisation by turning millions of lines of legacy code into requirements in hours, not years.

You’ll leave with practical patterns for agentic discovery, where it breaks down, and how to keep it honest.