====== Approved AI Use Cases ======

**Knowledge Bundle:** Public-Sector AI Readiness  
**Concept ID:** government:approved_ai_use_cases  
**Concept Type:** Use Case Library  
**Status:** Draft  
**Audience:** Public agencies, department leaders, program managers, IT teams  
**Steward:** OKF Expert  
**AI Use:** Human-reviewed  
**Last Reviewed:** June 2026

===== Purpose =====

Approved AI use cases identify practical ways public agencies may use AI while maintaining human accountability, protecting sensitive information, and preserving public trust.

AI should support staff judgment, not replace lawful authority, professional expertise, or public accountability.

===== Generally Appropriate Uses =====

  * Drafting routine communications for staff review
  * Summarizing long documents
  * Creating first drafts of meeting notes
  * Organizing internal knowledge bases
  * Improving search across policies and procedures
  * Creating training outlines and learning materials
  * Translating public information for staff review
  * Drafting public FAQs
  * Creating checklists, templates, and workflow guides
  * Assisting with internal research
  * Categorizing non-sensitive information
  * Supporting customer service staff with suggested responses

===== Uses Requiring Added Review =====

  * Public-facing chatbots
  * Grant-writing support
  * Procurement document drafting
  * Legal or policy research
  * Communications involving regulatory guidance
  * Content that could influence public decisions
  * Analysis involving demographic, financial, health, or protected data
  * Operational recommendations affecting staffing, funding, or service delivery

===== Uses That Should Not Be Approved Without Formal Review =====

  * Determining eligibility for benefits or services
  * Making employment decisions
  * Enforcement, investigation, or disciplinary decisions
  * Legal conclusions or legal advice
  * Safety-critical recommendations
  * Automated denial of public services
  * Facial recognition or biometric decision-making
  * Uploading confidential, protected, or personally identifiable information into unapproved systems
  * Releasing AI-generated content without appropriate human review

===== Approval Checklist =====

Before approving an AI use case, confirm:

  * The business problem is clearly defined.
  * A responsible owner has been assigned.
  * The proposed data use is understood.
  * Sensitive information will not be exposed.
  * Human review requirements are clear.
  * Risks have been assessed.
  * Required legal, privacy, security, and procurement reviews are complete.
  * Staff have received appropriate training.
  * Success measures and review dates are documented.

===== Related Concepts =====

  * [[government:ai_governance|AI Governance]]
  * [[government:ai_risk_assessment|AI Risk Assessment]]
  * [[government:ai_procurement|AI Procurement]]
  * [[government:staff_ai_training|Staff AI Training]]
  * [[government:ai_readiness|Return to Public-Sector AI Readiness]]