Journal of Pioneering Artificial Intelligence Research

Abstract

Autonomous or AI-assisted modernization of legacy monoliths is gaining attention as organizations seek to migrate to microservices without interrupting mission-critical services. However, decomposing a monolith safely remains difficult because the process introduces architectural risk, dependency uncertainty, data-consistency challenges, and operational failure modes that can directly affect availability. This paper presents a multivocal literature review of AI-driven approaches that support safe monolith decomposition and enable zero or near-zero downtime migration. Across the synthesized evidence, recurring patterns include runtime-informed partition discovery with explainable clustering, AI-guided dependency analysis to expose coupling risks, quality-driven decomposition approaches that treat migration as an optimization problem, AI-driven partitioning frameworks for service boundary design, and integrated toolchains that combine static and dynamic signals to recommend candidate microservices. Emerging directions further include large language model representations and contrastive learning to improve boundary quality, generative AI–driven code transformation for modernization, and systematic evidence mapping of AI methods and research gaps. Findings are presented as design propositions rather than universal claims, reflecting heterogeneity in evaluation settings and limited end-to-end migration validation under real operational constraints. The review contributes a structured pattern catalog linking AI capabilities (discovery, ranking, validation support) to safety goals (explainability, incremental rollout, rollback readiness), and it identifies future work needed to operationalize autonomy for continuous migration.