The Data Brain Architecture: How Artificial Intelligence, Master Data Management, and Enterprise Governance Are Transforming Modern Distributed Data Ecosystems
Year: 2026
Type: conference-paper
Abstract: Modern data warehousing environments increasingly deploy AI models in distributed data ecosystems for BI reporting, forecasting, customer intelligence, inventory analytics, and compliance monitoring. Yet these models frequently consume unstable features, duplicated master entities, fragmented lineage, and governance rules that remain outside real-time feature pipelines. This study develops and evaluates a Data Brain Architecture framework for improving AI feature reliability and governance compliance in cloud-based data warehousing pipelines. The framework integrates master data management, metadata contracts, entity-resolution constraints, lineage verification, feature-store monitoring, and policy-as-code into a single executable control layer. Using a design science research approach and matched quasi-experimental evaluation, conventional MDM-supported pipelines are compared with Data Brain-enforced pipelines. The evaluation shows that AI feature reliability increased from 78.5% to 93.9%, governance policy violation detection improved from 73.2% to 96.4%, lineage completeness improved from 72.6% to 95.8%, and master-data consistency increased from 79.8% to 94.6%. Duplicate entity propagation declined from 12.8% to 3.9%, prediction degradation dropped from 14.7% to 5.6%, and SHAP attribution drift reduced from 18.3% to 7.4%. Difference-in-differences results confirm significant treatment effects for feature reliability, governance compliance, lineage completeness, duplicate reduction, feature drift reduction, prediction stability, and explainability stability. The study contributes a theory-grounded and empirically evaluated architecture for AI-ready enterprise data governance in distributed data warehousing ecosystems.
Institutions: Accenture (Switzerland), Accenture (United States)
Language: English
Topic: Data Quality and Management
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7204132431