AI-Driven Data Governance Framework for Enterprises in Cloud Environments: Design, Implementation, and Enterprise Evaluation
Abstract
Challenges cloud-based enterprise data governance is encountering include data growth, regulatory changes, and the inflexibility of traditional rule-based systems. This systematic literature review, based on PRISMA guidelines, analyses 67 peer-reviewed papers (2019–2026) under three research questions related to AI-driven governance. Five weaknesses of a rule-based government are identified: rigidity of policies, poor scalability, semantic ambiguity, slow adaptation of policies, and weak anomaly detection. The benefits of AI techniques such as machine learning classifiers, large language models to extract policies, reinforcement learning to control access dynamically, and graph neural networks for federated lineage are measurable, with an 85-95% accuracy in detecting compliance and 38-70% reduction in manual effort. There is a common layered architecture (ingestion, intelligence, policy and enforcement) with policy-as-code, active metadata and federated lineage revealed in a comparative analysis. But there are still some stark contrasts, such as in the balance between the accuracy of a process and its explanation, between automation and auditability, and between cost reduction and assurance of compliance. The review presents a reference architecture and governance maturity model for enterprise adoption, and identifies open challenges such as model drift, multi-cloud consistency and legal and regulatory uncertainty, which require further research.
