A human-centered system for implementing AI policy in libraries
AI was already in the building. Staff were experimenting, patrons were asking questions, and vendor products were adding AI features. The missing piece was not another list of risks. It was a workable system for deciding what responsible use looked like in daily library work.
Robin Hastings designed and delivered a policy implementation method that connects durable principles to practical guardrails, staff training, tool governance, and scheduled review.
The challenge
Libraries needed to protect privacy, accuracy, professional judgment, and public trust without turning policy into a blanket prohibition. The approach also had to work across institutions with very different staffing, technical capacity, and approval processes.
The operating idea
Human defines the work. AI assists. Human reviews and owns the result.
That Human-AI-Human workflow becomes concrete through three safeguards:
- protect confidential information;
- verify facts, citations, calculations, and claims; and
- keep meaningful decisions and accountability with people.
What implementation included
- Purpose, scope, and guiding-principle drafting
- Approved, review-required, and prohibited use categories
- Patron-support and AI-literacy guidance
- Tool and vendor review questions
- Staff exercises tied to real policy decisions
- Related-policy review across privacy, collection development, acceptable use, programming, procurement, and records
- A 90-day path from inventory to rollout and reinforcement
Proof of execution
The method was delivered through a three-session California Libraries Learn workshop in August 2026, supported by an integrated workbook, three presentation decks, short and long policy templates, breakout drafting, and implementation planning. The registration record contains 14 registrations, with 10 marked as workshop attendees.
It was also adapted into a single-session ByWater Solutions webinar and later distilled into a 19-slide conference presentation. That progression demonstrates a framework designed to travel: from hands-on institutional work to executive briefing and professional education.
The evidence boundary
The workshop was designed to leave participating libraries with a policy draft and implementation plan. The available records do not verify how many institutions completed adoption or board approval, so this case study does not present those as measured outcomes.
The larger lesson
A policy PDF does not govern AI by itself. Governance becomes real when staff can recognize the boundary, know who owns the decision, practice the required judgment, and revisit the system as tools and community expectations change.
Download the full case study (PDF) to see the governance model, rollout timeline, training strategy, sample policy language, documented delivery signals, and lessons learned.
Robin Hastings helps libraries translate complicated technological change into operationally sustainable practice for human organizations.