Abstract
Large Language Models (LLM) offer significant potential for automating complex tasks across domains. This article presents a novel Enterprise Resource Planning (ERP) system that leverages LLM to fulfill customer requests by accessing the ERP database for real-time updates, modifications, and availability checks, enhancing interaction and efficiency. Using a graph-theoretic framework, the system supports stateful workflows with cycles, branching, and human-in-the-loop (HITL) interactions, allowing precise control over application flow. This approach redefines LLM deployment in ERP applications, providing enhanced automation and responsiveness in customer service.