> ## Documentation Index
> Fetch the complete documentation index at: https://docs.programmeinsights.co.uk/llms.txt
> Use this file to discover all available pages before exploring further.

# Generating Content

> Use AI-assisted content generation to draft remediation text for RED and AMBER findings.

## Before you start

* The finding must have a **RED** or **AMBER** rating -- content generation is not available for GREEN findings.
* You need **Reviewer** or **Admin** role on the assessment.
* Familiarise yourself with the finding's evidence gaps first via the Finding Detail view ([guide 07](/sops/07-finding-detail)).

## Steps

1. **Open content generation.** From a finding's detail view, click **Generate Content**. The content generation workspace loads.
   \[screenshot: 12-content-generation.png]

2. **Review the context assembly.** Before generating, the system assembles context from multiple sources:
   * **Finding details** -- the current RAG rating, evidence coverage percentage, and reviewer notes.
   * **Evidence gaps** -- the specific requirements where coverage is missing or insufficient.
   * **Criterion definition** -- the framework's formal description of what good evidence looks like.
   * **Framework terminology** -- the expected language conventions for the framework (e.g. IPA Green Book terms for HMT assessments).

3. **Configure tone and length.** Use the controls to adjust:
   * **Tone** -- formal, balanced, or concise.
   * **Length** -- short paragraph, standard section, or extended narrative.
     These settings influence how the AI drafts the remediation content.

4. **Generate the draft.** Click **Generate**. The AI drafts remediation content using the finding's cited evidence, the criterion definition, and the framework's expected language. Each field in the output shows a per-field confidence score indicating how well-supported the draft is by the source material.

5. **Review the output.** Read through the generated content carefully. The draft is structured to address the evidence gaps identified in the finding. Check that:
   * The language matches the framework's conventions.
   * The claims are supported by the cited evidence.
   * The remediation guidance is actionable.

6. **Choose your next action.** Three options are available:

   | Action         | What it does                                                                                     |
   | -------------- | ------------------------------------------------------------------------------------------------ |
   | **Revise**     | Edit the draft directly in the editor. Make targeted changes while keeping the AI's structure.   |
   | **Regenerate** | Discard the current draft and generate a fresh version. Useful if the tone or approach is wrong. |
   | **Accept**     | Save the content back to the finding. The accepted text becomes part of the finding record.      |

7. **Accept and save.** When satisfied, click **Accept**. The generated content is saved against the finding and appears in the finding detail view and in any reports that reference this criterion.

8. **Check generation analytics.** A generation analytics dashboard is available from the content generation workspace. It shows aggregate statistics across the assessment: how many findings have generated content, acceptance rates, and average confidence scores.

## What happens next

* Return to the finding to continue review: [Reviewing Findings](/sops/07-finding-detail).
* Continue working through results: [Understanding Results](/sops/06-results).
* See the reference diagram: `content-generation-flow.md`.

## Common questions

**Q: Can I generate content for a GREEN finding?**
A: No. Content generation targets remediation -- GREEN findings already have sufficient evidence and do not need remediation drafts.

**Q: Does accepting generated content change the finding's RAG rating?**
A: No. The RAG rating reflects the evidence in the assessed documents. Generated content is a remediation aid, not a substitute for submitting improved evidence. The rating updates only when a new assessment is run against updated documents.

**Q: What does the confidence score mean?**
A: Each field's confidence score (0--100%) indicates how much of the draft is grounded in cited evidence versus inferred by the AI. Higher scores mean stronger source backing. Review low-confidence fields with extra care.