AS/400 Modernization Assessment → POC → Pilot

A large healthcare provider running independent IBM i/AS/400 Db2 environments across multiple country locations needed a safer path off fragmented legacy data structures without disrupting registration and downstream clinical operations. The program chose a staged model: standardize and centralize DDS-based assets first, then create a governed foundation for broader modernization. GenAI and Agentic AI accelerated discovery, dependency mapping, rule reconstruction, and test planning, while deterministic tooling and human review preserved traceability, rollback readiness, and audit confidence.

The estate was stable, but stability had become expensive. Separate Db2 iSeries databases across sites created semantic drift, duplicate structures, and inconsistent naming, slowing integration, analytics, and change delivery. Registration workflows, identity creation, and surrounding batch dependencies still had to remain dependable while leaders faced rising platform costs, tighter compliance expectations, shrinking legacy skills, and stronger interoperability demands. Every modernization decision therefore had to protect patient access, claims continuity, data privacy, and service levels during peak periods and audit review.

STAR*M GenAI & Agentic AI-Led Discovery with Controlled Execution

The assessment avoided two common mistakes: a risky big-bang rewrite and a simple lift-and-shift that would preserve technical debt. Instead, the team used automated refactoring, tool-assisted conversion with validation controls, to convert DDS structures into Db2 SQL, unify schemas across two sites, centralize them in an on-premises test environment, and migrate the validated target to Amazon Aurora PostgreSQL.

The POC covered multiple patient-facing modules and DDS files, with controlled test-data migration, one-time reverse replication, and formal acceptance gates. In parallel, GenAI generated inventory summaries, mapping drafts, and test hypotheses, while Agentic AI orchestrated multi-step discovery tasks under engineering oversight. That created a credible runway for later API enablement and, if required, Java-based service modernization.

What the POC & Pilot De-risked

The immediate benefit was measurable risk removal. Within a six-week POC and a twelve-week Pilot window, leadership gained a governed inventory, a phased migration plan, validated conversion deliverables, defined rollback mechanics, and clear responsibility boundaries across project management, architecture, cloud, migration, and development teams.

Based on this scoped pattern, leaders could conservatively model lower run costs, faster release preparation, improved schema consistency, stronger developer productivity, and better readiness for HL7/FHIR integration, analytics, and AI use cases as planning outcomes rather than booked production results. Just as important, business testing, compliance validation, and care-continuity safeguards remained explicit rather than assumed.

Healthcare leaders should treat AI as an Accelerator for Discovery, not as a substitute for controls. Start with evidence-led inventorying. Standardize data semantics before moving code. Use phased slices with rollback paths and explicit acceptance criteria. Keep SMEs, testers, and security teams engaged from day one.

Prioritize platforms that support APIs, analytics, and future Java or microservices evolution. The winners modernize in governed increments, then use the cleaner data and service estate to support interoperability, predictive operations, and tighter cost control.