Overview and Implementation Guide
About this document set
This is the entry point to the Relyance AI Use Cases documentation. Start here, then branch out to the guide that matches your role:
| Document | Audience | What's in it |
|---|---|---|
| Overview and Implementation Guide (this doc) | Everyone | What an AI use case is, how it connects to your inventory, assessments, policies, and risks, and a rollout plan |
| Admin Configuration Guide | AI governance, privacy, and security teams | Creating use cases, attaching components, working the use case drawer, similar use cases, and assessments |
| Change Management and Lifecycle | AI governance reviewers and use case owners | Approving a use case, how changes are detected and scored, change requests, and lifecycle stages |
| Legal and Compliance Guide | Legal, compliance, and audit teams | The AI governance obligations a use case record supports and what it demonstrates - no UI steps |
| Audit Log and Reporting | Compliance, auditors, governance leads | The Use Cases dashboard and grid, the audit log, and the compliance package export |
| FAQ: AI Use Cases | Prospects, new customers, sales and CS | Common questions in plain language |
AI Use Cases is in Beta and is available to select customers. Contact your account team for access. Change management - approval, change detection, and lifecycle stages - is enabled separately, and is described as available "where enabled" throughout these guides.
What is an AI use case?
An AI use case is a documented AI system in your organization: what it does, who owns it, how risky it is, and what it is built from. It is the unit regulators and frameworks reason about. The EU AI Act, the NIST AI Risk Management Framework, and ISO/IEC 42001 all attach obligations to an AI system and its intended purpose, not to a model in isolation.
Relyance already discovers the building blocks of your AI estate - models, AI agents, MCP servers, and the vendors behind them - and lists them under Observe → AI. A use case ties a set of those building blocks to a business purpose, so you can govern the system as a whole:
- Components - the models, AI agents, MCP servers, and SaaS vendors the use case runs on, picked from your inventory.
- Business context - name, description, department, owner, risk level, and your own custom fields.
- Derived posture - governance status, usage status, GenAI flag, data use, and sensitive data exposure, all rolled up from the components and kept current as scans run.
- Governance evidence - the assessments that cover it, the policies that apply to its components, the risks and issues raised against it, attachments, and an audit trail.
Components are shared, not owned. One model or agent can be part of many use cases, and a change to that component shows up on every use case that uses it.
Where to find it
Go to Observe → AI and select the Use Cases tab. Where AI Use Cases is enabled, Use Cases is the first tab and the page opens on it, next to Models, AI agents, and MCP servers.
The tab has three charts across the top - Use Cases by Risk Level, Sensitive Data Exposure, and Governance Coverage - and a grid of every use case below them. Click a row to open the use case drawer. See The Use Cases dashboard.
How the pieces fit together
Getting use cases in. You can create a use case in three ways:
- Describe it with AI - write a sentence or two about what the AI does, optionally attach a design doc or spec, and Relyance drafts the use case: name, description, department, owner, risk level, custom field values, and matching components from your inventory. You review the draft before anything is saved.
- Fill it in manually - walk a four-step wizard: Basics, Components, Details, and Review.
- From the inventory - open a model, AI agent, or MCP server, and in its Associated Use Cases section start a new use case with that component already in it.
See Creating a use case.
Avoiding duplicates. While you create a use case, Relyance compares it with your existing ones. If it looks like one you already have, you can compare the two side by side and add to the existing use case instead. Each use case also keeps a Similar Use Cases card, from which you can reuse a match's assessment answers, components, and review owners. See Similar use cases.
Assessing. When you create a use case, Relyance can start two assessments against it straight away: the AI Use Case Intake assessment, a short triage questionnaire, and the full AI Governance and Risk Assessment. Any other assessment you run against the use case, or against one of its models or vendors, appears on its Assessments card. See Assessments on a use case.
Monitoring. Nothing evaluates a use case directly. Instead, Policy Center policies evaluate its components, and the use case shows the result: the Policies card lists each policy that applies with its status, and the Risks & Issues card lists the policy violations against its components alongside the Risk Register risks raised from its assessments.
Approving and watching for change. Where change management is enabled, approving a use case records a baseline. Relyance then compares every scan against that baseline, scores each difference, accepts low-risk changes on its own with an explanation, and holds higher-risk changes for a person to review - which can reopen the affected assessment questions. See Change Management and Lifecycle.
Proving it. Every create, edit, component change, approval, and lifecycle move is recorded in the use case's audit log, and Export Compliance Package bundles the use case report, model reports, assessment reports, risks, issues, and audit log into one ZIP for auditors. See Audit Log and Reporting.
Who uses it
- An AI governance team keeps a single register of every AI system in production or development, with an owner and risk level for each, and runs the intake and risk assessment before launch.
- A privacy team sees which use cases process sensitive data, and which personal-data categories each one touches, without asking engineering.
- A security team sees which policies are failing on the agents and MCP servers behind a use case, and gets a review when an agent gains a destructive tool or becomes over-privileged.
- Legal and compliance export a per-system evidence pack for an auditor or regulator.
Implementation plan
Prerequisites. Use cases are built from what Relyance has discovered, so connect the sources that feed the AI inventory first - source code repositories, cloud accounts, and observability tools. See Source Code Integration and Infrastructure Integration. You can create a use case with no components and attach them later, but its governance status, data use, and sensitive data exposure stay empty until it has some.
Step 1 - Agree what counts as a use case. Decide the level at which you catalog AI systems. A use case should map to one business purpose with one accountable owner - for example, "Support reply drafting" rather than "All of Customer Support's AI" or "The GPT-4o deployment".
Step 2 - Set up your fields. Make sure your Department values and any custom fields for AI use cases are in place in Field Customizations, so every use case is captured the same way. The AI draft can also suggest custom fields worth adding.
Step 3 - Seed the catalog. Start with the AI systems you already know about. Describing each with AI is the fastest route; for systems you discover in the inventory first, start the use case from the component's drawer.
Step 4 - Assess. Leave both assessments ticked when you create each use case, assign them, and work them in Assessments. Raise risks from the completed assessments into the Risk Register.
Step 5 - Register and approve. Register each component on its own tab so the use case's governance status reaches Registered, then approve the use case where change management is enabled.
Step 6 - Operate. Review the Changes tab when a use case moves to Re-review, keep owners and review dates current, and export a compliance package when an auditor asks.
Permissions
Anyone who can see Observe → AI can view use cases. Creating, editing, deleting, attaching components, approving, accepting changes, and moving a use case through its lifecycle need edit access to the platform.