We combined AI-assisted reverse engineering, specification-driven development and human engineering expertise to create a controlled pathway from legacy code to modern software.

A purpose-built AI reverse-engineering capability mapped the legacy estate across 27 business domains and 1,000+ source and custom files. It identified relevant legacy logic, extracted business rules and generated language-agnostic pseudocode, with rules traced directly back to their original source code.

This created a reliable knowledge layer between the legacy system and the new platform, allowing AI to work from verified business behaviour rather than assumptions.

From there, each capability followed a repeatable AI-assisted delivery pipeline:

Legacy source → User story → Specification → AI-generated implementation → Automated testing → Engineer review → Deployment

Rather than attempting a large-scale rewrite, the team first delivered a complete end-to-end walking skeleton, taking a core process from customer creation and scheduled services through to scheduling, run sheets, batches and invoicing.

AI handled the heavy lifting of analysis and implementation, while engineers retained ownership of architecture, decisions and quality. Every change was validated against acceptance criteria, automated tests and the original legacy behaviour before deployment.

The result was a modern, cloud-based foundation that could be continuously deployed and used as the platform for the wider modernisation programme.

‍

The project demonstrated that AI-amplified engineering can significantly compress the time and team size required to modernise complex legacy systems, without compromising engineering rigour.

By the numbers:

• 709K+ lines of legacy code analysed
• 653 legacy screens mapped
• 244 legacy programs
• 27 business domains reverse-engineered
• 1,000+ source files analysed
• 4 maximum team members
• 3 months from standing start to deployed software
• 1,323 automated tests at the July snapshot

Work that would traditionally take months with a significantly larger team was compressed into weeks, and in some cases days.

The project delivered a working cloud-based business workflow, a reusable modern application foundation and a structured knowledge base of the legacy system.

Most importantly, it proved a repeatable AI-assisted modernisation model that can now be applied across the wider estate. AI accelerates the work. Engineers own the judgement. Specifications define the behaviour. Tests provide the proof.

‍