Transitional-State Architectures for Hybrid Mainframe to Cloud Modernization With AI-Derived Code Intelligence

Most mainframe modernization programs fail or stall not in the target state, but during transitional phases where legacy and cloud workloads coexist. In parallel, artificial intelligence (AI)-enabled code understanding tools increasingly produce rich metadata—call graphs, data flows, domain clusters, and inferred business capabilities—that are rarely exploited to design transitional architectures and to evaluate them empirically. This article proposes a reference framework for AI-informed transitional-state architectures in mainframe-to-cloud modernization. Using multiwave evidence from two multiyear programs (anonymized), we show how AI-generated dependency graphs and domain clusters can guide strangler patterns, carve-out sequencing, and network segmentation. We define a key performance indicator set for transitional-state health that extends beyond on-time/on-budget delivery to include cutover risk, batch window compliance, incident rates, and cross-domain coupling. The contributions include: 1) taxonomy of AI-informed transitional patterns; 2) decision heuristics that map code-level insights to architectural moves; and 3) reusable evaluation blueprint.