Before You Build the Chatbot: Matching AI to the Library Problem

Abstract

Abstract We argue, in this column, for a problem-first approach to generative AI adoption in libraries. Using a searchable FAQ system and archival finding aids as contrasting cases, we distinguish between bounded retrieval tasks and exploratory discovery. The FAQ case shows how generative AI can improve metadata without replacing conventional search, while the archival case suggests a stronger use for conversational AI: helping researchers identify relevant collections, boxes, and relationships across descriptive records. We conclude that libraries should consider adopting conversational AI when it adds meaningful capability, improves discoverability, or enables work that existing systems or human workflows cannot already perform efficiently or effectively.