Slide 1 – From AI anxiety to operational practice
Implementing human-centered AI policy in libraries
AI was already present. The work was to make responsible judgment visible, teachable, and repeatable.
Slide 2 – The gap was operational
Libraries were balancing four realities:
- Staff and patrons were already using AI.
- Vendor products were adding AI features.
- Shared expectations often lagged behind use.
- A rigid tool list would become obsolete quickly.
The design question: How do we create room to act without losing privacy, accuracy, or accountability?
Slide 3 – Human-AI-Human
- Define – A person sets the purpose, constraints, and acceptable inputs.
- Assist – AI produces bounded options, drafts, patterns, or summaries.
- Review and own – A person checks privacy, facts, bias, tone, accessibility, and consequences.
The final human step is where professional responsibility lives.
Slide 4 – Policy is a system
Policy sets purpose, scope, principles, and boundaries.
Procedures make tool review, documentation, and escalation usable.
Training gives staff practice applying the policy to real decisions.
Review keeps the system current as tools and expectations change.
Slide 5 – From template to practice
Session 1: Define
Map current use and draft purpose, scope, and principles.
Session 2: Decide
Draft approved and prohibited uses, patron guidance, data protection, and tool governance.
Session 3: Deploy
Peer-review language, plan staff training, connect related policies, and name the next approval and rollout steps.
Slide 6 – Documented delivery signals
- 3 workshop sessions x 90 minutes
- 14 registrations; 10 marked as workshop attendees
- Integrated workbook, 3 decks, 2 policy templates, and implementation tools
- Framework adapted to a separate ByWater Solutions webinar
- Material distilled into a 19-slide conference presentation
Evidence boundary: institutional adoption and board approval were not verified in the available records.
Slide 7 – What this work demonstrates
Policy expertise + systems design + adult learning + change management + delivery
The result is not only a policy template. It is a repeatable method for turning public-service values into daily decisions about AI.