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

# Call Analysis

> Extract insights, summarize conversations, and automate post-call workflows

## Overview

eigi.ai's Call Analysis feature automatically processes every conversation to extract valuable insights, generate summaries, and trigger automated workflows. Turn unstructured conversations into structured, actionable data.

***

## Enabling Call Analysis

### Configuration

<Steps>
  <Step title="Navigate to Analysis Tab">
    Open your agent and select the "Analysis" tab
  </Step>

  <Step title="Enable Analysis">
    Toggle "Call Analysis Enabled" to activate the feature
  </Step>

  <Step title="Configure Schema">
    Define what data you want to extract from calls
  </Step>

  <Step title="Set Up Notifications">
    Optionally configure alerts based on analysis results
  </Step>
</Steps>

***

## Analysis Schema

### Define What to Extract

Create a custom schema that tells the AI exactly what information to pull from conversations:

<CardGroup cols={2}>
  <Card title="Text Fields" icon="font">
    Extract text data like names, reasons, feedback
  </Card>

  <Card title="Number Fields" icon="hashtag">
    Capture numeric values like ratings, quantities, scores
  </Card>

  <Card title="Boolean Fields" icon="toggle-on">
    Yes/No determinations like "interested", "resolved"
  </Card>

  <Card title="Choice Fields" icon="list">
    Categorical values from predefined options
  </Card>
</CardGroup>

### Example Schema

| Field                | Type    | Description                   |
| -------------------- | ------- | ----------------------------- |
| `customer_sentiment` | Choice  | positive, neutral, negative   |
| `issue_category`     | Text    | Type of issue discussed       |
| `resolution_status`  | Boolean | Was the issue resolved?       |
| `satisfaction_score` | Number  | 1-10 rating if provided       |
| `follow_up_needed`   | Boolean | Does customer need follow-up? |
| `key_topics`         | Text    | Main topics discussed         |
| `action_items`       | Text    | Next steps identified         |

***

## Analysis Prompt

### Customizing Analysis

Write specific instructions for how calls should be analyzed:

```text theme={null}
Analyze this customer service conversation and extract:

1. Customer Sentiment: Determine if the customer's overall sentiment
   was positive, neutral, or negative based on their tone and words.

2. Issue Category: Identify the main reason for the call
   (billing, technical, general inquiry, complaint, etc.)

3. Resolution: Did the agent successfully resolve the customer's issue?
   Answer true or false.

4. Satisfaction Score: If the customer expressed satisfaction or
   gave a rating, capture it on a 1-10 scale. Otherwise, estimate
   based on the conversation tone.

5. Follow-up Needed: Based on the conversation, does this customer
   need any follow-up action? Answer true or false.

6. Summary: Provide a 2-3 sentence summary of the call.
```

***

## Call Recording

### Capture Every Interaction

Enable recording for compliance, training, or quality assurance:

<AccordionGroup>
  <Accordion title="Recording Settings" icon="circle-dot">
    * **Enable/Disable**: Toggle recording per agent - **Automatic Start**:
      Recording begins when call connects - **Secure Storage**: Encrypted storage
      with access controls
  </Accordion>

  <Accordion title="Playback" icon="play">
    * Review recordings directly in the dashboard - Synchronized with
      conversation transcript - Adjustable playback speed
  </Accordion>

  <Accordion title="Retention" icon="clock">
    * Configurable retention periods - Automatic cleanup of old recordings -
      Export before deletion if needed
  </Accordion>
</AccordionGroup>

***

## Video Analysis

### For Video-Enabled Agents

When using video avatars, additional analysis capabilities are available:

* **Engagement Detection**: Measure user attention and engagement
* **Visual Context**: Capture context from screen shares
* **Interaction Quality**: Assess video call quality and experience

<Info>
  Video analysis is available for agents using HeyGen or Tavus video avatars.
</Info>

***

## Structured Data Extraction

### Automatic Data Capture

The AI extracts structured data from natural conversations:

**Example Conversation:**

```
Agent: "May I have your email address for the confirmation?"
Customer: "Sure, it's john.smith@email.com"
Agent: "And what date works best for you?"
Customer: "How about next Tuesday at 2pm?"
```

**Extracted Data:**

```json theme={null}
{
  "email": "john.smith@email.com",
  "preferred_date": "Tuesday",
  "preferred_time": "2:00 PM"
}
```

***

## Analysis Results

### Viewing Results

Access analysis data for each conversation:

<Steps>
  <Step title="Open Call History">
    Navigate to Call History in the dashboard
  </Step>

  <Step title="Select Conversation">Click on any completed conversation</Step>

  <Step title="View Analysis">
    See the Analysis tab with all extracted data
  </Step>

  <Step title="Export">Download data in JSON or CSV format</Step>
</Steps>

### Data Fields

Each analyzed call includes:

| Field             | Content                        |
| ----------------- | ------------------------------ |
| **Transcript**    | Full conversation text         |
| **Summary**       | AI-generated call summary      |
| **Duration**      | Call length                    |
| **Outcome**       | Success/failure/transfer       |
| **Custom Fields** | All schema-defined extractions |
| **Timestamps**    | Start, end, key moments        |

***

## Notifications

### Alert Configuration

Trigger notifications based on analysis results:

<CardGroup cols={2}>
  <Card title="Email Alerts" icon="envelope">
    Send email when specific conditions are met
  </Card>

  <Card title="Trigger Fields" icon="bolt">
    Alert based on any extracted data field
  </Card>

  <Card title="Threshold Rules" icon="sliders">
    Notify when values exceed or fall below thresholds
  </Card>

  <Card title="Instant Delivery" icon="paper-plane">
    Receive alerts immediately after call analysis completes
  </Card>
</CardGroup>

### Example Triggers

| Trigger            | Condition                     | Action                  |
| ------------------ | ----------------------------- | ----------------------- |
| Hot Lead           | `lead_score > 80`             | Email sales team        |
| Negative Sentiment | `sentiment = negative`        | Alert support manager   |
| Escalation Needed  | `escalation_requested = true` | Notify supervisor       |
| High Value         | `deal_value > 10000`          | Alert account executive |

***

## Use Cases

### Quality Assurance

* Monitor agent performance across all calls
* Identify training opportunities
* Track resolution rates and customer satisfaction

### Sales Intelligence

* Score and prioritize leads automatically
* Extract buying signals and objections
* Track conversion metrics

### Compliance

* Ensure required disclosures are made
* Verify identity confirmation steps
* Audit conversation quality

### Customer Insights

* Identify common pain points
* Track feature requests
* Measure sentiment trends

***

## Best Practices

<Tip>
  **Start Simple**: Begin with a few key fields and expand as you learn what's
  valuable.
</Tip>

<Tip>
  **Be Specific**: Clear, specific schema fields yield more accurate
  extractions.
</Tip>

<Tip>
  **Test Your Schema**: Run analysis on test calls to verify fields extract
  correctly.
</Tip>

<Warning>
  Ensure your analysis practices comply with privacy regulations and disclose
  recording where required.
</Warning>

***

## Troubleshooting

<AccordionGroup>
  <Accordion title="Analysis not running">
    * Verify "Call Analysis Enabled" is toggled on - Check that you have an
      analysis prompt configured - Ensure the call completed successfully
  </Accordion>

  <Accordion title="Fields extracting incorrectly">
    * Review your analysis prompt for clarity - Make field definitions more
      specific - Test with sample conversations
  </Accordion>

  <Accordion title="Missing data">
    * Some fields may be empty if data wasn't discussed - Add "N/A" or default
      value instructions to your prompt
  </Accordion>
</AccordionGroup>