AI in Action at ETRI Library

Abstract

What can a research library do with AI when it has no in-house development team? At ETRI Library, six projects began with everyday frustrations rather than technology: borrowing stopped when staff were away, researchers needed help with manuscript similarity checking, metadata work was slow, and more than 50,000 publications remained difficult to find. The library responded by transforming an ordinary office PC into a self-service checkout, return, and pickup kiosk, while installing nine low-cost displays for meeting-room information. It simplified manuscript similarity checking through institutional SSO and evaluated an LLM for recovering existing abstracts from paper PDFs. Evidence-checking and RPA workflows are also being developed or piloted for research-output records. Beyond daily services, AI supported efforts to improve Google Scholar discoverability and gave librarians a stronger voice in planning a new library space. The six cases range from operational services to evaluated pilots and early prototypes, but share a common approach: librarians defined the problem, used AI for bounded and reviewable tasks, tested results against evidence and local requirements, and involved vendors and specialists when needed. Human judgement remained central. Together, the cases show that AI can help libraries do more than automate existing work. It can support original service design, reshape workflows, and open new possibilities for the library’s role—even without extensive technical resources.