In many IBM i environments, mission-critical business processes continue to rely on traditional 5250 green-screen interfaces, leading to challenges with knowledge transfer, skills shortages, and modernization pressures. This session explores a technical pattern for achieving secure, 100% reproducible AI-driven automation of real 5250-based workflows. By recording and analyzing actual operator interactions (keystrokes, screens, validations), these processes can be transformed into structured, documented representations—such as JSON-based microservice schemas—that capture exact input/output behaviors, validation rules, and flow logic without exposing underlying RPG/COBOL source code. These representations serve dual purposes: (1) enabling local AI models to execute the processes identically to a human operator (ensuring zero deviation and full auditability), and (2) acting as a living knowledge base for natural-language querying (e.g., "Explain the customer creation steps" or "Where do validation failures commonly occur?"). The approach leverages local AI deployments (no cloud data leakage), fine-grained security controls (e.g., JWT tokens and role-based access), and automated documentation generation to support onboarding, process optimization, and regression testing. Attendees will gain insights into the architectural building blocks, recording techniques, schema design principles, integration with IBM i security models, and practical considerations for maintaining fidelity in AI-assisted execution.This session provides technical guidance on bridging legacy IBM i processes with modern AI capabilities while preserving trust, compliance, and operational integrity in an era of workforce transitions.
