Where Can AI Help Interlibrary Loan—and Where Has It Not?

p strongAI interest has reached interlibrary loan, but operational adoption remains early: even obvious-looking applications such as title searching had not yet entered routine use in the surveyed libraries./strong/p p Primary Research Group's 55-page study of 33 institutions found that no participant was using AI for title searches. At lower-tuition institutions&mdashthose charging less than $7,500&mdashstaff reported a mean of about 23 minutes per week using ChatGPT, showing experimentation without broad workflow integration./p p Expectations also differed by institution. Public libraries were somewhat more optimistic about productivity, with some expecting AI to improve performance by more than 50%. Private institutions were more cautious. That range matters because resource sharing depends on reliability: a confident but incorrect match, citation or routing decision can waste staff time across more than one library./p p The sensible starting point is low-risk assistance around the transaction rather than unsupervised fulfillment. AI may help summarize lending policies, draft patron s, classify exception reasons or analyze cancellation patterns. Bibliographic identification, copyright decisions and final routing still require trusted data and accountable review./p p ILL managers should measure a pilot against existing turnaround time, fill rate, staff touches, error rate and patron satisfaction. The goal is not to AI into every step it is to locate repetitive work where assistance reduces delay without weakening accuracy. Early benchmark data is especially valuable now, before isolated experiments harden into unexamined practice./p p  /p