Discuss an assessment RESOURCES
Understand the work.
Explore the outputs.
A practical guide to the records, relationships and decisions that make up an AI governance workspace.
Go to the glossary ↘WORKING PREVIEWS
Four ways to use the record.
AI registry and risk observations
Systems, purposes, owners, model context and recorded risk considerations.
Explore the previewEXAMPLE OUTPUTEvidence and control gap pack
Bounded observations, supporting records and proposed review decisions.
Explore the previewEXAMPLE OUTPUTExecutive AI risk brief
A leadership summary with a scoped registry and source appendix.
Explore the previewEXAMPLE OUTPUTOngoing workspace and action register
Assigned work, status, priority, due dates and the evidence needed to close it.
Explore the previewA SHARED LANGUAGE
Different records.
Different jobs.
- AI system
A business application or operational workflow using AI.
IN THE DEMONSTRATIONCommercial Underwriting Assistant.
- Use case
The business work and decision the system supports.
IN THE DEMONSTRATIONSummarize a submission for a human underwriter.
- Model
A specific predictive or language model used by a system.
IN THE DEMONSTRATIONAn explicitly fictional hosted language model; not inferred from the application’s name.
- Deployment
Where and how a system/model version runs.
IN THE DEMONSTRATIONAn authored example environment and hosting region, or “Not documented.”
- Requirement
An obligation or expectation whose source and applicability must be identified.
IN THE DEMONSTRATIONA clearly fictional oversight requirement in the demonstration.
- Policy
The organization’s stated rule or commitment.
IN THE DEMONSTRATIONHuman review is required before a defined decision.
- Control
An assigned activity intended to put a policy or requirement into practice.
IN THE DEMONSTRATIONA reviewer checks and records a decision before work advances.
- Evidence
A record supporting a specific assertion, scope and period.
IN THE DEMONSTRATIONA procedure describes the control; a review event may show an instance of its operation.
- Knowledge view
Entities and typed relationships: what exists and how it is connected.
IN THE DEMONSTRATIONSystem → model → provider; policy → control → evidence.
- Context view
Recorded rationale, actors, source versions, conditions and changes.
IN THE DEMONSTRATIONWhy a mapping was accepted, who reviewed it, and what changed afterward.
A policy statement is not proof that a control operated. Accepting a source assertion is not approval of an AI use or acceptance of its risk. A “second brain” can help retrieve and connect information; it does not replace these distinctions or the customer review process.