A Compliance-Focused Retrieval-Augmented AI System for Pharmacy Policy Assistance

Abstract

Pharmacy operations depend on detailed regulatory and procedural policies that are often spread across lengthy documents, making them difficult to search and interpret consistently. This paper presents a compliance-focused Retrieval-Augmented Generation (RAG) system that enables pharmacists to query internal policies in natural language and receive concise, citation-grounded responses. The solution integrates Azure Cognitive Search with an enterprise LLM, supported by structured prompts and a HIPAA-aligned architecture.Twelve chunking configurations were tested, with a 400-character chunk and 100-character overlap achieving the best balance of similarity (0.775), correctness (97.9%), and token efficiency. In a pilot with eight pharmacists, the system reduced policy lookup time by over 80% and received a satisfaction rating of 4.6/5.Overall, the system demonstrates that RAG can deliver reliable, traceable, and efficient policy guidance in regulated pharmacy environments, while offering a practical and repeatable framework for organizations adopting AI-assisted policy interpretation.

Ähnliche Arbeiten