// CHALLENGE
When patching is no longer enough.
The application supported active rail operations but its ageing technology stack had reached its limits. Critical security vulnerabilities could no longer be resolved without replacing the underlying platform.
This wasn't a straightforward rewrite. The existing architecture tightly coupled Keystone with MongoDB 5.0. Moving forward meant migrating unstructured data into a new structured model while preserving data integrity and compatibility with the application already used in signal boxes.
The system had to change.
The operation couldn't stop.
// APPROACH
Rethink the system. And how we build it.
Rather than applying AI to the product, we applied it to the engineering process itself.
Structured AI workflows supported analysis, planning, implementation and testing throughout the rewrite. Our engineers remained in control of architecture and quality while automation removed much of the work surrounding traditional development.
- End-to-End orchestration: Without relying on a Product Owner or designer, a multi-agent framework powered ticket creation, sprint planning, analysis, code generation, testing, and verification.
- Human-in-the-loop oversight: Expert developers provided guidance throughout, validating code quality and architectural integrity.
- Data migration: Unstructured MongoDB data was transformed into a modern structured model while preserving integrity.
- Seamless transition: By replacing manual coding and administrative overhead with systematic AI flows, we eliminated traditional bottlenecks and transformed complex unstructured data into a modern, structured database—all while preserving the familiar user experience and maintaining live compatibility with the operational mobile app.
// Result
Three times faster. Zero downtime.
The legacy platform was replaced with a modern system while rail operations continued uninterrupted.
Critical vulnerabilities were removed. Existing data was migrated into the new model. The mobile application remained operational throughout the transition.
AI-assisted engineering helped deliver the rewrite up to 3x faster than the traditional development approach, while establishing a repeatable way of working for future engineering projects.
"By embedding structured AI workflows into our process, we rebuild a critical system faster without lowering the standard."
- Danny Matthijs, AI Lead at SumZero
// Impact
By replacing manual coding and administrative overhead with automated AI workflows, we fundamentally shifted how fast legacy modernization can be delivered:
- 3x faster delivery: the platform rewrite was completed in a fraction of the expected development time.
- Critical vulnerabilities resolved: unpatchable legacy infrastructure was replaced rather than worked around.
- Zero operational downtime: active mobile users continued working throughout the transition.
- >60% less development overhead: AI workflows reduced the coordination and manual work surrounding the engineering process.
