# Admin Configuration Guide

*Audience: AI governance, privacy, and security teams who create and maintain AI use cases.*

This guide covers creating AI use cases, attaching their components, and working the use case drawer. For approval, change detection, and lifecycle stages, see [Change Management and Lifecycle](/docs/ai-use-cases/change-management-and-lifecycle/). For what an AI use case is, start with the [Overview and Implementation Guide](/docs/ai-use-cases/overview-and-implementation-guide/).

:::info
AI Use Cases is in Beta and is available to select customers. Contact your account team for access. Every action on this page that changes a use case needs edit access to the platform.
:::

## Creating a use case

Go to **Observe → AI**, select the **Use Cases** tab, and click **Add Use Case**. The **Add Use Case** panel opens and asks **What will this AI do?** From here you either describe the use case and let Relyance draft it, or fill it in yourself.

### Describe it with AI

1. In the text box, describe the use case in a sentence or two - what it does, the model or vendor it uses, the data it works with, and who owns it. For example: *Support wants a GPT-4o agent that drafts replies to Zendesk tickets using order history. Maya owns it.*
2. Optionally, drop or attach a design doc or spec as extra context: PDF, Word, or text files, up to 3 files of 5 MB each.
3. Click **Describe with AI**, or press ⌘ + Enter (Ctrl + Enter on Windows).

Relyance reads your description, fills in the name, department, owner, and risk level, analyzes it for your custom fields, and matches components from your inventory - models, agents, MCP servers, and vendors. This usually takes under 30 seconds. The panel then opens on the **Review** step with everything filled in.

On the review step, check what the draft did:

- **Fields filled from your description** are marked with the Relyance AI mark. The draft also explains the risk level it chose.
- **Names it couldn't match** - for example, an owner who isn't a Relyance user, or a component that isn't in your inventory - are called out so you can pick the right one or add the component later.
- **Suggested custom fields** are fields the draft thinks are worth tracking for every use case. Tick a suggestion to add that field to every use case; hover the ✦ to see why it was suggested.

Click **Edit** on any row to go back to that step and change it. If the draft fails, or you'd rather not use it, click **Fill in manually**. The files you attached are kept on the use case either way.

### Fill it in manually

Click **Fill in manually** to walk the wizard. The stepper at the top shows the four steps, and you can jump back to any step you've already reached.

1. **Basics** - enter a **Name** and **Description** (both required). Optionally choose a **Department** and a **Use Case Owner**, set the **Risk Level** (**High**, **Medium**, or **Low**), and add **Attachments**: PDF, Office, text, or image files, up to 10 files of 25 MB each.
2. **Components** - answer **What does it run on?** by picking models, AI agents, MCP servers, and vendors from your inventory. This step is optional; you can add components later.
3. **Details** - fill in your organization's custom fields for AI use cases. All are optional.
4. **Review** - check everything, then choose which assessments to start (see below).

### Starting assessments when you create a use case

The **Review** step lists two assessments, both ticked by default:

- **Start the intake assessment** - creates *\<name\> Initial Assessment* from the **AI Use Case Intake** template: a short triage questionnaire covering purpose and scope, data, impact, controls, and a triage outcome, including whether the use case needs a full AI risk assessment.
- **Start the AI Governance and Risk Assessment** - creates *\<name\> AI Governance and Risk Assessment* from the **Artificial Intelligence Governance and Risk Assessment** template: the detailed review, with sections on AI system classification, data governance, transparency, human oversight, recordkeeping, and the additional sections required for high-risk AI systems.

Untick either one to skip it. If a similar use case already has a completed intake, the review step also offers **Start from the answers of \<name\>**, which copies its answers into the new intake for you to review.

Click **Add Use Case** to save. The use case is created first; if an assessment, attachment, or custom field value then fails to save, you see a warning, and the use case itself is kept.

### Creating a use case from the inventory

You can also start from a component you've found in the inventory:

1. Open a model, AI agent, or MCP server from its tab on **Observe → AI**, or from the [Inventory](/docs/inventory/inventory-management-guide/).
2. In its **Associated Use Cases** section, click **Add to Use Case**.
3. Pick an existing use case to attach the component to, or click **New Use Case** to open the **Add Use Case** panel with this component already selected.

The new use case opens on top of the component's drawer, so you can go straight back to the component when you're done.

### If the use case looks like one you already have

While you create a use case, Relyance compares it with your existing ones. If there's a close match, a **Looks like** banner shows it with its match percentage. Click **Compare** to see the two side by side, with their purpose, components, and data, and shared rows highlighted. If it's the same AI system, click **Add to it instead** to add your components to the existing use case rather than creating a duplicate.

## The use case drawer

Click a row in the **Use Cases** grid to open its drawer. The header shows the use case's name, a **GenAI** tag when any component is an agent, an MCP server, or a generative model, and an **Export Compliance Package** button (see [Exporting a compliance package](/docs/ai-use-cases/audit-log-and-reporting/#exporting-a-compliance-package)).

Below the header, a row of fields summarizes the use case:

| Field | What it shows | Editable? |
|---|---|---|
| **Lifecycle** | The use case's lifecycle stage - shown where change management is enabled | Through the **Approval** card |
| **Governance Status** | **Registered** only when every attached model, agent, and MCP server is registered; otherwise **Unregistered**. Hover it to see which components are holding it back | No - it's derived from the components |
| **Usage Status** | Rolled up from the components' own usage status | Yes - your value overrides the rollup |
| **Department** | The owning department | Yes |
| **Risk Level** | **High**, **Medium**, or **Low** | Yes |
| **Business Owner** | The person accountable for the use case | Yes |

The drawer has an **Overview** tab, a **Changes** tab where change management is enabled, and an **Audit Logs** tab. The **Overview** tab holds the cards below.

### Governance Status card

The governance record for the use case: **Governance and Security Review Owner(s)**, **Registered On**, **Last Reviewed**, **Next Review**, and **Summary Notes** - the reviewer's narrative, in rich text. Click a value to edit it. Where change management is enabled, approving the use case records the review and schedules the next one for you; see [Review dates](/docs/ai-use-cases/change-management-and-lifecycle/#review-dates).

### Key Details card

- **Description** - click to edit.
- **Data Use** - every data category the components process, most sensitive first. Derived from the components, so it can't be edited here.
- **Products** - the products the components belong to, also derived.
- **Team**, **Deployment Environment**, and **Data Residency** - free-form fields you record yourself; each shows *Not recorded* until set.
- **Created**, **Last Updated**, and **Last Modified By** - **Last Updated** reflects the last change a person made to the use case or its components, not component scans.

### Risks & Issues card

Two views, switched with the **Risks** and **Issues** toggle:

- **Risks** - the [Risk Register](/docs/risk-register/overview-and-implementation-guide/) risks that apply to this use case, with their status, inherent risk, owner, and whether they're monitored. Risks raised from the use case's assessments are linked to it automatically. See [From an assessment](/docs/risk-register/admin-configuration-guide/#from-an-assessment).
- **Issues** - the policy violations against this use case's components, worst first, one row per policy and component, with severity, status, and when it was detected. Counts match the default [Issues list](/docs/policy-center/issues-and-ai-review/#what-the-issues-list-shows-by-default).

### Components card

The models, AI agents, MCP servers, and SaaS vendors the use case is built from, with the governance and usage status each one contributes. Click a component's name to open its own drawer on top of the use case.

- **To attach a component**, click **Add Component** and pick it from your inventory.
- **To detach a component**, use its remove action and confirm. The component itself is untouched; only its membership in this use case is removed.

A component that has since disappeared from your inventory stays on the list marked **No longer in the inventory**, so it isn't dropped silently.

:::tip
Vendors don't count toward governance status or usage status. A use case with only vendors attached, or with nothing attached, shows as **Unregistered**.
:::

### Similar Use Cases card

See [Similar use cases](#similar-use-cases) below.

### Custom fields card

Your organization's custom fields for AI use cases, with this use case's values. Manage the fields themselves in [Field Customizations](/docs/other-settings/field-customizations/).

### Policies card

The [Policy Center](/docs/policy-center/overview-and-implementation-guide/) policies that apply to this use case's components, each with its category, a status, and when it was last evaluated:

| Status | Meaning |
|---|---|
| **Violation** | At least one component has an open or in-progress violation of the policy. The row shows how many are open out of the total |
| **Needs Review** | The policy has fired on a component, but every violation is resolved or closed |
| **Compliant** | The policy applies and nothing has violated it |

Policies evaluate components, not the use case directly, so this card is a view of how the use case's building blocks are doing.

### Assessments card

See [Assessments on a use case](#assessments-on-a-use-case) below.

### Attachments card

Files kept on the use case for reviewers and assessments - design docs, specs, and the brief it was drafted from. Add PDF, Office, text, or image files, up to 10 files of 25 MB each.

## Similar use cases

Relyance looks for use cases that resemble each other when a use case is created, and stores what it finds on both use cases. The **Similar Use Cases** card shows each match with its similarity percentage, description, and the reasons it was matched.

- **Re-compute** checks again against your current use cases. The card shows when it was last checked.
- **Compare** opens the two use cases side by side. If they aren't related, click **Not similar** there to dismiss the match.

From **Compare**, you can **Reuse from \<name\>** - copy what the match already has into this use case. Choose any of:

- **Assessment answers** - copied into this use case's intake. Answers only fill questions that aren't answered yet here, and you still review them.
- **Its components** - their policies come with them.
- **Review owners**.

Then click **Reuse selected**. If the match has a completed intake, the card also prompts you to **Reuse answers** directly.

## Assessments on a use case

The **Assessments** card lists two kinds of assessment:

- Assessments whose subject is this use case - including the intake and AI Governance and Risk Assessment started when you created it.
- Assessments that reach it through an attached vendor or model.

Each row shows the assessment's type, status, assignee, and due date, and opens the assessment. You can also create any assessment against a use case from Assessments; see [Creating assessments](/docs/assessments/admin-configuration-guide/#creating-assessments).

When an assessment about the use case is complete, raise risks from it into the Risk Register; they're linked back to the use case automatically. See [Raising risks from completed assessments](/docs/assessments/admin-configuration-guide/#raising-risks-from-completed-assessments).

## Editing and deleting a use case

Edit fields in place in the drawer: click a value, change it, and save. Every edit is recorded in the use case's audit log.

To delete a use case, open the **⋯** menu on its row in the grid and choose **Delete**. If any assessments name the use case as their subject, the confirmation tells you how many - they're kept, but are left without a subject. Deleting can't be undone.

## Configuration checklist

- [ ] AI Use Cases is enabled for your tenant (and change management, if you want approvals and change detection).
- [ ] Integrations that feed the AI inventory are connected, so components are available to attach.
- [ ] **Department** values and custom fields for AI use cases are set up in Field Customizations.
- [ ] Each AI system is cataloged as a use case with an owner and a risk level.
- [ ] Components are attached, and each model, agent, and MCP server is registered.
- [ ] The intake and AI Governance and Risk Assessment are assigned for each use case.
- [ ] Risks from completed assessments are raised into the Risk Register.
