Getting Started
Key Concepts
Core terminology and mental models used throughout the AI Council Toolkit.
Terminology
| Term | Definition |
|---|---|
| AI Council | The central cross-functional governance body responsible for AI oversight |
| Champion | An individual embedded in a business unit or team who acts as liaison between the council and delivery teams |
| Tiering | The process of classifying AI use cases by risk level to determine the appropriate review pathway |
| Intake | The process by which new AI use cases are registered and assessed for the first time |
| Triage | The routing step that determines whether a use case is self-serve, champion-reviewed, specialist-reviewed, or council-reviewed |
| Impact assessment | A structured evaluation of an AI system's potential effects on people, the organization, and society |
| Model card | A standardized document describing a machine learning model's intended use, performance, and limitations |
| Datasheet | A standardized document describing a dataset's provenance, composition, and intended use |
| AI inventory | A register of all AI systems in use or development across the organization |
Mental Models
Layered Governance
The best AI governance is not a single committee. It is a layered system:
- Executive sponsor. Sets tone, allocates resources, accountable to the board.
- AI Council. Central policy, triage, and review body.
- Specialist reviewers. Security, privacy, legal, domain experts called in for specific reviews.
- Champions. Embedded in teams, first point of contact, handle low-risk guidance.
- Delivery teams. Build and operate AI systems, responsible for completing assessments and following policy.
Front-Door Review + Ongoing Monitoring
Governance is not a gate you pass through once. Effective councils run two loops:
- Front-door review. New use cases are assessed before deployment.
- Ongoing monitoring. Deployed systems are periodically reviewed, incidents are tracked, and policies are refreshed.
Risk-Based Routing
Not every use case needs the same level of scrutiny. A tiering system routes cases to the right level:
- Tier 1 (Low risk). Self-serve with templates and guidelines.
- Tier 2 (Medium risk). Champion review with lightweight assessment.
- Tier 3 (High risk). Full council review with impact assessment.
- Tier 4 (Prohibited / needs escalation). Escalated to executive sponsor or blocked.