IGeLU 2026

Conference

This presentation explores the critical paradigm shift from traditional search engines to advanced reasoning systems, redefining library intelligence through Agentic AI. Tracing the 73-year evolution of information retrieval from early computing in 1953, we highlight how the focus has moved from simple lexical matching and document ranking to generative, multi-turn conversational reasoning. Today, information is nothing without the ability to reason, synthesize, and understand complex, ambiguous user intents.

While naive chatbots are prone to severe academic hallucinations, licensing violations, and black-box logic, we introduce the robust concept of Library-Grade Agentic AI. This approach utilizes coordinated multi-agent architectures—where specialized agents (such as Policy, Retrieval, and Reasoning agents) collaborate dynamically. These intelligent workflows ensure precise, proactive knowledge discovery and automated metadata cataloging. They strictly adhere to institutional constraints, access rights, and verifiable sources, comprehensively transforming both front-end user research and back-end library operations.

However, deploying autonomous AI introduces profound challenges. We examine these risks across four dimensions: user risks like intellectual property leakage, filter bubbles, and hidden hallucinations; library staff risks such as indirect prompt injections and cascading systemic failures; leadership challenges including massive resource exhaustion and severe vendor lock-in; and a concerning transparency gap among developers regarding AI safety testing. To mitigate these threats, the presentation details four communication architectures (Centralized, Shared Message Pool, Decentralized, Layered) and emphasizes the non-negotiable necessity of Human-in-the-Loop governance. Ultimately, the librarian's role is elevating from a passive searcher to an active Agent Architect. Our mission is not to compete with AI, but to govern it, ensuring it remains accountable to verified knowledge.

IGeLU 2026 theme: "Pearls of Technology: Think deeply, explore widely"