Secure Enterprise AI Agent for Procurement Insights across SAP and Ariba Systems using RAG and Apigee X

In modern enterprise environments, procurement and accounts payable (AP) operations are fragmented across disparate systems, including SAP ECC, SAP Ariba, and business warehouses (BW), leading to siloed financial data. Accessing accurate and timely information, such as purchase order (PO) or invoice status, typically requires specialized expertise, creating significant operational delays, and an over-reliance on support teams. This paper presents a Natural-Language Procurement Data Assistant, an enterprise-grade AI Agent designed to provide real-time contextual answers to procurement queries via a secure, scalable, and API-driven framework. The core solution architecture integrates Google Apigee X for API management and security, Lang Chain/Lang Graph for multistep agentic reasoning, and a Retrieval-Augmented Generation (RAG) pipeline. This pipeline translates natural-language inputs into precise SQL queries executed against the Enterprise Data Platform (EDP) hosted on Google BigQuery as the single source of truth consolidating all SAP and Ariba data. The agent is deployed within a containerized Google Kubernetes Engine (GKE) environment, interfacing with an internal Gen-AI front-end for end-user interaction. By strictly grounding all the responses in the EDP, the system guarantees accuracy, consistency, and auditability across financial operations. Empirical validation across live procurement scenarios demonstrates a high-performance system achieving