This paper presents the design, implementation, and evaluation behind a Large Language Model (LLM) agent that chats with an industrial production-grade ERP system. The agent is capable of interpreting natural language queries and translating them into executable SQL statements, leveraging open-weight LLMs. A novel dual-agent architecture combining reasoning and critique stages was proposed to improve query generation reliability.
@article{arxiv.2507.23429,
title = {Chatting with your ERP: A Recipe},
author = {Jorge Ruiz Gómez and Lidia Andrés Susinos and Jorge Alamo Olivé and Sonia Rey Osorno and Manuel Luis Gonzalez Hernández},
journal= {arXiv preprint arXiv:2507.23429},
year = {2025}
}
Comments
11 pages, includes 3 tables summarizing schema and model performance. Submitted on July 31, 2025. Targets integration of LLM agents with ERP systems using open-weight models and Ollama deployment