[2603.20815] GMPilot: An Expert AI Agent For FDA cGMP Compliance

Computer Science > Artificial Intelligence arXiv:2603.20815 (cs) [Submitted on 21 Mar 2026] Title:GMPilot: An Expert AI Agent For FDA cGMP Compliance Authors:Xiaohan Wang, Nan Zhang, Sulene Han, Keguang Tang, Lei Xu, Zhiping Li, Xiue (Sue)Liu, Xiaomei Han View a PDF of the paper titled GMPilot: An Expert AI Agent For FDA cGMP Compliance, by Xiaohan Wang and 7 other authors View PDF Abstract:The pharmaceutical industry is facing challenges with quality management such as high costs of compliance, slow responses and disjointed knowledge. This paper presents GMPilot, a domain-specific AI agent that is designed to support FDA cGMP compliance. GMPilot is based on a curated knowledge base of regulations and historical inspection observations and uses Retrieval-Augmented Generation (RAG) and Reasoning-Acting (ReAct) frameworks to provide real-time and traceable decision support to the quality professionals. In a simulated inspection scenario, GMPilot shows how it can improve the responsiveness and professionalism of quality professionals by providing structured knowledge retrieval and verifiable regulatory and case-based support. Although GMPilot lacks in the aspect of regulatory scope and model interpretability, it is a viable avenue of improving quality management decision-making in the pharmaceutical sector using intelligent approaches and an example of specialized application of AI in highly regulated sectors. Comments: Subjects: Artificial Intelligence (cs.AI) Cite as: arXiv...