Public libraries · Staff readiness · Patron privacy
What should library staff know before offering public AI programs?
Library staff need three things before a public AI program: a written boundary for what staff will and will not do with AI on a patron's behalf, a patron privacy rule that covers uploaded documents and personal details, and a response for inaccurate output. Staff readiness comes first. Public programming comes after.
Staff training is not public AI literacy programming
A patron class explains what the tools do. Staff training decides what a person at the desk is authorized to do when a patron asks for help with a résumé, a benefits form, an immigration document, or a medical question. Those are different problems and they need different sessions.
What library staff practice in session
- The desk script for AI help requests, including the refusal case.
- Patron privacy: what may be uploaded, and what never leaves the patron's own hands.
- Verifying output before it is passed to a patron as information.
- Referral and escalation for legal, medical, benefits, and immigration questions.
- Documenting a decision so the next shift handles it the same way.
- Branch consistency across locations and part-time staff.
Delivery options for systems
- Single staff-day session for one branch or system.
- Multi-branch delivery with a shared boundary document.
- Train-the-trainer so internal staff repeat the program without ongoing external cost.
- Sponsor debrief with directors and branch-services leaders.
Accessibility and language
Materials are built with accessibility in mind, and bilingual English/Spanish delivery is available when included in scope. Content is educational and operational. It is not legal, clinical, or financial advice.
Questions buyers ask
Direct answers
- How can libraries protect patron privacy during AI assistance?
- Keep patron documents and identifying details out of staff accounts and shared tools, help at the patron's own device where possible, and record nothing beyond what circulation policy already allows.
- What should librarians tell patrons before they upload documents to an AI tool?
- That anything uploaded may be retained by the vendor, that identifying details should be removed first, and that the library cannot guarantee the accuracy of the result.
- How should libraries respond when AI produces inaccurate information?
- Treat it as an unverified source. Say plainly that it may be wrong, verify against a known reference, and document the correction so other branches respond the same way.
- What does responsible AI training for librarians include?
- Approved tools, a desk script, patron privacy rules, verification practice, escalation paths, and one documented change per branch with an owner and a date.
Next step
Discuss staff-facing library support.
Bring the number of branches, the staff count, and the programming date you are working toward. We confirm format and fit in twenty minutes.
Contact question: what audience and outcome does your initiative need to support?