Service
Legacy-to-modern with AI-assisted refactoring, translation, and re-platforming — without freezing the business while you do it.
Frequently Asked
COBOL, PL/I, RPG, and Natural on mainframe and AS/400. Java EE monoliths, .NET Framework, classic ASP, PowerBuilder, Oracle Forms. SAP ABAP customisations. Older PHP, Perl, VB6, and a fair amount of undocumented stored-procedure logic. If a vendor stopped supporting it, we've probably worked on it.
It helps measurably in three places: comprehension (reverse-engineering business rules from undocumented code), translation (initial pass at code conversion that humans then verify), and test generation (creating behavioural test suites against legacy that act as the safety net for the new system). It does not replace architectural judgement.
Strangler-fig pattern with capability-by-capability cutover. New system stands up alongside legacy, traffic shifts incrementally behind feature flags, and we keep a rollback path live until the new system has earned trust against production load.
12–36 months for a programme; 8–16 weeks for a discovery-and-roadmap engagement that gets you to a credible plan, business case, and first cutover candidate.
Common. We capture institutional knowledge through working sessions, code archaeology, and behavioural test generation — the goal is that the system survives the people who originally built it, regardless of when modernisation completes.
Strangler-Fig Cutover
What we build
We take systems that the business depends on but the engineering team is afraid of — and turn them into systems that are safe to change. Mainframe modernisation, monolith decomposition, re-platforming to cloud, replacing end-of-life middleware, and the careful kind of refactoring that uses AI as a tool, not a magic wand.
How we approach it
- Comprehension first. Code archaeology — what does the system actually do? We use LLMs to reverse-engineer business rules from undocumented code, then validate every output against the runtime. The deliverable is a current-state behavioural specification, not a wishlist.
- Capability map and cutover plan. We identify the strangler boundaries — which capabilities can be lifted out independently, in what order, with what dependencies. The cutover sequence is the architecture.
- Test net before transformation. Behavioural tests against legacy become the contract the new system must satisfy. AI accelerates test generation; humans certify the assertions.
- Incremental cutover. New capability stands up, traffic routes to it behind a flag, parity is monitored in production, rollback is one config change away. We move when the data agrees we can move.
- Decommission discipline. Modernisation is not done until the legacy system is off. We plan the decommission from day one and track it as a deliverable.
Where AI helps
- Comprehension — extracting business rules from COBOL/RPG/PL/I, ABAP, or PowerBuilder where the documentation is the code.
- Translation — first-pass code conversion across language families, with human verification on every output.
- Test generation — behavioural tests, edge-case discovery, regression suites against legacy.
- Refactoring at scale — applying consistent patterns across thousands of files faster than a team could review by hand.
- Documentation — turning the modernised system into something the next team can actually read.
Where AI does not help
Architectural judgement. Domain modelling. Negotiating which capabilities the business can live without during cutover. Deciding when to rewrite and when to wrap. These remain human decisions, made with engineers who have done this before.
Capabilities
- Mainframe modernisation (COBOL, PL/I, RPG, Natural; CICS, IMS, DB2).
- Monolith decomposition and microservices migration.
- Cloud re-platforming (AWS, Azure, GCP) and lift-and-reshape strategies.
- Database migration: Oracle to Postgres, DB2 to managed services, schema rationalisation.
- Middleware replacement (TIBCO, MQ, legacy ESBs).
- API enablement and integration modernisation.
- Engineering enablement: CI/CD, observability, test culture, on-call.
Why Eldridge Morgan
Modernisation is a multi-year contact sport. We staff it with engineers who have done it before, use AI where it earns its place, and refuse the shortcut of a big-bang rewrite. The system runs the business; we treat it accordingly.