Abstract
Abstract
The rapid growth of scientific literature and other unstructured materials has made agricultural decision support increasingly complex. Students, extension agents, inspectors, and private sector researchers increasingly rely on large language models (LLM). These models can produce incomplete, overly cautious, or erroneous responses. We outline how LLM offers a possible opportunity for the animal science discipline, describe what LLM do well and where they fall short, and explain how retrieval-augmented generation reduces fabrication and improves relevance. We present a case study that builds a compact index from the United States Department of Agriculture Animal and Plant Health Inspection Service (APHIS) inspection narratives for Agricultural Research Service bird facilities and show how those passages ground concise answers for nutritionists in that audience. We propose an integrated system in which frontline decision makers gain access to LLM while accessing sound information through a lens of subject-matter expertise.