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

# Document Generation

> How to generate AI-powered documents from your calls and signals

BuildBetter's document generation transforms your recordings, signals, and data into structured business documents using AI. Follow these steps to create professional documents from your workspace content.

## Generation Process

<Steps>
  <Step title="Select Template" stepNumber={1}>
    Choose from available templates (PRD, user story, meeting summary, etc.).

    <ul>
      <li>Each template has predefined structure and sections</li>
      <li>Templates include default generation settings and data source preferences</li>
      <li>Preview templates to see example outputs</li>
    </ul>
  </Step>

  <Step title="Add Data Sources" stepNumber={2}>
    Select the content to analyze - you can mix and match sources:

    <ul>
      <li><b>Calls</b>: Individual recordings with transcripts</li>
      <li><b>Folders</b>: Collections of related content</li>
      <li><b>Signals</b>: Specific extracted insights</li>
      <li><b>Datasets</b>: Organized data collections</li>
    </ul>
  </Step>

  <Step title="Configure Generation Options" stepNumber={3}>
    Set document title, focus, and generation parameters:

    <b>Generation Style</b>:

    <ul>
      <li><b>Focused Analysis</b> (one-shot): Best for small datasets (under 100 signals or under 10 calls). Simple all-at-once processing.</li>
      <li><b>Theme Discovery</b> (divide-and-conquer): Best for large datasets (over 100 signals or over 10 calls). Advanced pattern recognition across large datasets. Not suitable for counting or quantitative analysis.</li>
    </ul>

    <b>Data Format</b>:

    <ul>
      <li><b>Full Context</b> (qualitative): Uses complete transcripts and conversations. Best for narrative documents and detailed analysis. Only available when using calls or folders (not datasets/extractions).</li>
      <li><b>Key Insights</b> (extractions): Uses pre-processed signals and extractions. Best for summaries and reports focused on specific insights. Less effective with 20+ sources due to context overload.</li>
    </ul>

    <b>Visibility</b>:

    <ul>
      <li><b>Workspace</b>: Anyone in your workspace can view</li>
      <li><b>Private</b>: Only you can view (default)</li>
    </ul>
  </Step>

  <Step title="Generate & Review" stepNumber={4}>
    Click Generate and monitor progress. Once complete:

    <ul>
      <li>Review the generated content</li>
      <li>Edit directly in the document editor</li>
      <li>Regenerate if needed with different settings</li>
      <li>Share with your team when ready</li>
    </ul>
  </Step>
</Steps>

<Note>
  The document title and focus (description) fields are optional but highly recommended. The more detail you provide, the more accurate and targeted the document will be.
</Note>

## Generation Strategies Explained

### Focused Analysis (One-Shot)

**Best for**: Small datasets (under 100 signals or under 10 calls)

* Processes all content at once
* Ideal for focused, cohesive documents
* All context fits together naturally
* Faster generation for small datasets

**Use when**:

* Creating meeting summaries from 1-5 calls
* Generating PRDs from focused customer feedback
* Building documents from specific signal sets

### Theme Discovery (Divide-and-Conquer)

**Best for**: Large datasets (over 100 signals or over 10 calls)

* Breaks down content to identify patterns
* Advanced pattern recognition across sources
* Identifies emerging themes automatically
* Synthesizes insights from large volumes

**Not suitable for**:

* Counting or quantitative analysis
* Exact tallying of specific items

**Use when**:

* Analyzing trends across many calls
* Creating reports from large signal sets
* Identifying patterns in extensive datasets

## Data Source Options

### Full Context (Qualitative)

Uses complete transcripts and conversations.

**Best for**:

* Narrative documents
* Detailed analysis
* When you need the full story with all nuances
* Documents requiring direct quotes

**Requirements**:

* Only available with calls or folders as sources
* Not available when using datasets or extraction-only sources

### Key Insights (Extractions)

Uses pre-processed signals and extractions.

**Best for**:

* Summaries and reports focused on specific insights
* When signals have already been extracted
* Structured insights over narrative

**Limitation**:

* Less effective with 20+ sources due to context overload
* Best with focused, curated signal sets

<Info>
  The Data Format option adapts based on your source selection. If you select datasets or extractions only, "Full Context" won't be available.
</Info>

## Document Actions

Once your document is generated, you can:

### Edit & Refine

* Edit content directly using the rich text editor
* Auto-save keeps your changes safe
* Regenerate the entire document if needed

### Share & Collaborate

* Copy document link to share with teammates
* Set visibility to Workspace for team access
* Use document chat to discuss and iterate

### Copy Content

* **Copy Document**: Copies document content to clipboard (both HTML and plain text formats)
* **Copy Link**: Shares the document URL for team access

### Delete

* Remove documents you no longer need
* Confirmation required to prevent accidental deletion

## Document Chat

Every document includes an integrated chat panel:

* Ask questions about the document content
* Get clarification on generated insights
* Request edits or additions
* Iterate on content with AI assistance

**To access**:

1. Open any document
2. Click the chat icon in the top-right
3. Start a new conversation or continue existing threads
4. Document context is automatically included

<Note>
  Document chat threads are saved and can be resumed later. You can have multiple chat threads per document.
</Note>

## Source Selection Tips

<CardGroup cols={2}>
  <Card title="Single Call" icon="phone">
    Use for meeting summaries, detailed call analysis, or focused PRDs from specific customer conversations.
  </Card>

  <Card title="Multiple Sources" icon="layer-group">
    Use for trend analysis, comprehensive reports, or documents synthesizing insights from various inputs.
  </Card>

  <Card title="Folders" icon="folder">
    Use to analyze entire collections of related content at once.
  </Card>

  <Card title="Signals" icon="sparkles">
    Use when you want documents focused on specific extracted insights.
  </Card>
</CardGroup>

## Generation Progress

During generation, you'll see:

* Real-time status updates
* Stage messages
* Percent completion when available
* Progress counters for multi-step generation
* Processing indicators
* Completion notifications
* Error alerts if issues occur

If live progress events are missed, BuildBetter falls back to durable status tracking so the document can still show reliable completion or error state.

Documents typically generate within 30 seconds to 2 minutes, depending on:

* Amount of source data
* Generation strategy chosen
* Current system load

## Best Practices

<Check>
  **Provide clear titles and focus**: Detailed titles and descriptions help AI understand your intent
</Check>

<Check>
  **Choose appropriate generation strategy**: Use Focused Analysis for small sets, Theme Discovery for large datasets
</Check>

<Check>
  **Match data source to need**: Use Full Context for narrative, Key Insights for summaries
</Check>

<Check>
  **Start with fewer sources**: Test with 2-3 calls before generating from 50+ sources
</Check>

<Check>
  **Use document chat for iteration**: Rather than regenerating entirely, use chat to request specific edits
</Check>

<Warning>
  Always review AI-generated content for accuracy and completeness before sharing. The AI provides intelligent synthesis but may occasionally miss nuances or make errors.
</Warning>

## Troubleshooting

<AccordionGroup>
  <Accordion title="Generation Fails or Takes Too Long">
    * Reduce the number of sources
    * Try switching generation strategy
    * Ensure all sources are fully processed
    * Check that calls have completed transcription
  </Accordion>

  <Accordion title="Poor Quality Output">
    * Add more detail to title and focus fields
    * Try different data source option (Full Context vs Key Insights)
    * Ensure source content is relevant to document goal
    * Use document chat to request specific improvements
  </Accordion>

  <Accordion title="Missing Content">
    * Verify all selected sources contain relevant data
    * Check source permissions and access
    * Try Full Context mode if using Key Insights
    * Add more source material if output is thin
  </Accordion>
</AccordionGroup>

## Limitations

<Tip>
  Use the Copy Document feature to paste content into Google Docs, Notion, or other tools. Content is copied with both HTML formatting and plain text for compatibility.
</Tip>