Metadata-Driven Data Governance for Enterprise AI Readiness: A Scalable Framework for Lineage, Discovery, and Regulatory Compliance

Abstract

Artificial intelligence (AI) investments in enterprise organisations are more and more finding themselves with a foundational paradox: superior, accessible, and regulated data assets are the primary drivers of AI outcomes, but systematic management of data assets is not seen in most data deployed environments. This research proposes a five-modules architecture for a metadata-driven data governance solution that can assist in navigating enterprise readiness for AI, using data lineage, data semantic cataloguing, intelligent discovery and AI regulatory compliance automation. Based on the structured survey, the respondents were 250 people in 6 industry sectors, which were selected for the empirical study using analysis approach of mixed methods consisting of Pearson correlation analysis, multiple linear regression, structural equation modeling, one-way ANOVA and chi-square contingency testing. The first correlations that stood out were with metadata catalogue maturity (β = 0.34, p < .001), data lineage coverage depth (β = 0.29, p < .001), and compliance policy automation (β = 0.26, p < .001), all of which made a statistically significant contribution to enterprise AI readiness. The model exhibited a good fit (CFI = 0.96 and RMSEA = 0.048) and explained 74% of the variance in AI readiness outcomes (R² = 0.74) with metadata governance constructs. A one-way ANOVA across the five data governance maturity levels showed that the variance across organisations is accounted for by the governance maturity of AI readiness, F(4, 245) = 162.7, p < .001, η² = 0.726. The data asset discovery time reduction had a 99.9% improvement score, metadata completeness had a score of 81.3 percentage points and a score of 52.4% on the regulatory audit was found to be improved. The suggested framework offers a practitioner-oriented, theory-informed approach to making data governance a strategic levers for enterprise AI at scale, an operational model.