Mr. Padmanabham Venkiteela

Secure Retrieval-Augmented AI Agent for Natural-Language Procurement Insights Across SAP and Ariba Systems

Abstract:

Contemporary enterprise financial functions, especially procurement and accounts payable (AP), struggle significantly with data fragmentation. Core financial transaction details are scattered across disparate platforms such as SAP ECC, SAP Ariba, and Business Warehouse (BW) making it slow and difficult for teams to acquire a complete, accurate understanding of a purchase order or invoice. This fragmented landscape necessitates reliance on specialized SAP expertise, which results in persistent operational delays and a dependence on support teams. To overcome this challenge, we introduce the Natural-Language Procurement Data Assistant, an enterprise-grade AI Agent engineered to provide real-time, contextually accurate answers to procurement queries. The solution is founded on a secure, scalable, and API-first architecture, with its core innovation being a Retrieval-Augmented Generation (RAG) pipeline utilizing LangChain/LangGraph for advanced, multi-step reasoning. This agent directly translates natural language questions into optimized SQL queries, which are then executed against a centralized Enterprise Data Platform (EDP), hosted on Google BigQuery. By establishing the EDP as the single source of truth that unifies all SAP and Ariba data, the system guarantees accurate, consistent, and auditable responses. The agent is deployed within a containerized Google Kubernetes Engine (GKE) environment, interfacing with the Trellix Hopper Gen AI front-end. Crucially, Google Apigee X is integrated to provide the necessary API management and security, enforcing enterprise-grade authentication, throttling, and access policies. Empirical testing on live procurement scenarios demonstrates significant operational benefits: the system achieves near-perfect accuracy (98–100 percent), maintains a rapid average query latency of less than 4 seconds, and reduces the AP team workload by a substantial 60 percent. This integrated architecture serves as a robust reference model for deploying secure, explainable, and compliant AI agents within complex enterprise procurement ecosystems, effectively transforming conversational AI into a tool for real-time operational intelligence.

Profile:

Mr. Padmanabham Venkiteela is a highly accomplished Senior Enterprise Integration Architect with over 18 years of experience designing, leading, and implementing robust, enterprise-scale integration solutions across diverse industries. Renowned for his strategic vision and hands-on expertise, he excels in delivering scalable, high-performing integration architectures using modern patterns—synchronous, asynchronous, real-time, batch, and API-led connectivity.

Mr. Venkiteela brings deep technical proficiency in a broad spectrum of leading integration and API technologies, including Boomi (Integrations & API Management), Apigee X, Oracle SOA Suite (11g/12c), Oracle Cloud Integrations, Oracle Service Bus, SAP BTP Integration Suite, Informatica Cloud, and Google Application Integration. His strong command of CI/CD pipelines using tools like Jenkins ensures rapid, secure, and reliable solution deployment.

A multi-cloud certified professional (GCP, AWS, Oracle Cloud) and a recipient of multiple MVP awards, he is recognized for consistently delivering impactful, enterprise-grade integration strategies—including data center migrations and cloud transitions to AWS and GCP.

Complementing his integration expertise, Mr. Venkiteela holds a Post Graduate Program in Artificial Intelligence and Machine Learning, with a specialization in Generative AI and agent development using Google Vertex AI—equipping him with the foresight to align emerging AI technologies with enterprise integration needs.

With a rare blend of architecture acumen, hands-on delivery, and future-focused AI skills, Mr. Venkiteela continues to be a strategic asset for organizations looking to modernize and scale their digital ecosystems.