AI-Driven Secure Data Provisioning for ERP Quality Environments in the Cloud: Architecture, Implementation, and Impact on Automated Financial Reconciliation

Abstract

Modern enterprises migrating to cloud-hosted Enterprise Resource Planning (ERP) platforms face critical challenges in maintaining secure, compliant nonproduction environments while preserving operational realism for testing and reconciliation activities. Traditional data refresh practices expose sensitive personal, financial, and proprietary information, creating significant regulatory and audit risks. This work presents a comprehensive architecture integrating machine learning classification, deterministic tokenization, and ratio-preserving cost transformations to enable secure data provisioning for ERP quality environments. The framework addresses sensitive field detection across financial modules, maintains referential integrity through consistent identifier tokenization, and preserves ledger relationships essential for automated reconciliation processes. Material Master costing data receives specialized treatment through controlled distortion techniques that maintain cost ordering and valuation logic while protecting proprietary manufacturing intelligence. Evaluation demonstrates high masking coverage, preserved integration stability across finance and supply chain modules, and maintained reconciliation accuracy. The architecture supports enterprise modernization initiatives, including cloud migration, clean core strategies, and compliance strengthening, while enabling realistic testing scenarios for financial automation and analytics capabilities.