---
title: "Score a risk with AI"
canonical: "https://support.appfire.com/space/RM/3459121540/Score%20a%20risk%20with%20AI"
format: markdown
---
Appfire LiveRisk uses Forge LLM to analyze a risk and generate a proposed risk score.

Open a risk work item and go to the **Risk analysis** panel. If the panel isn't visible, select **View app actions** below the work item summary, then select **Risk analysis**.

## Before you begin

- To use AI-assisted risk actions, you need **Manage LiveRisk Risk Register** and the Jira permissions required to view and edit the risk work item.
- The **LiveRisk Manager** and **LiveRisk Analyst** roles have **Manage LiveRisk Risk Register** by default.
- Provide enough information about the risk for LiveRisk to assess its impact and probability. If more context is needed, LiveRisk can ask refinement questions.

## Generate an AI estimate

In **Risk analysis**, expand **AI estimate**.

1. Select **AI estimate** to generate an estimate. If an estimate already exists, select **Re-run estimate**.
2. Wait for the estimate to complete. The draft shows the proposed impact and probability values, the corresponding tier, and a rationale for the estimate.
3. Review the result. When LiveRisk has enough context, the estimate shows:
  - The proposed Impact × Probability score and corresponding risk tier
  - A rationale for **Likelihood** and **Impact**
  - A confidence level for each
4. If LiveRisk doesn't have enough information to calculate a score, the estimate shows **Not scored** and explains what information is missing. You can:
  - Select **Refine** to answer the model's questions.
  - Add more information to the risk and select **Re-run estimate**.
5. To refine the estimate, select **Refine**, answer the displayed questions, then select **Send answers**.
6. Review the updated estimate and select **Accept** to apply it. **Accept** is unavailable when the estimate does not contain a score.

What LiveRisk saves depends on the risk's current status:

| **Status when you accept** | **What gets saved** |
| --- | --- |
| Identified, Evaluation | Inherent score and Current score |
| Treatment, Monitoring | Current score only |
| Re-evaluation | Current score only, unless the Inherent score is missing or the estimate exceeds it |

> 📝 There is no **Reject** action. To discard a draft, leave it unaccepted or use **Re-run estimate** to generate another estimate.

> ℹ️ Saved refinement answers remain available as risk owner context for future estimates. You can remove them from **Saved answers**.

## Review previous iterations

The **AI estimate** section includes previous estimates so you can review how the model has assessed the risk over time and as its details and linked signals have changed. Expand a previous estimate to review its scoring summary.

## Review the AI conversation

LiveRisk keeps the conversation used to generate an AI estimate so you can review the context and interactions behind the result. Click the conversation bubble icon in the right corner to reveal the conversation.

For long-running conversations, LiveRisk compresses the conversation and continues it in a new segment instead of truncating the earlier content.

A continued segment is identified by **continued** in its tab label. Its transcript begins with a notice that the conversation was continued from an earlier segment and compressed into the summary shown.

Transcript entries can be labeled by type, including:

- **System**
- **Tool**
- **Assistant**
- **Risk Owner**
- **Context**
- **Context change**
- **Conversation summary**
- **Instruction**
- **JSON reminder**