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

# Actor Insights

> How to configure Insights for your Vocalis AI Actor. Define the structured data points — scores, text, and charts — that get extracted from every conversation.

Insights are the structured data points that Vocalis extracts from each conversation. Instead of just giving you a raw transcript, the AI Actor pulls out the specific information you care about — scores, key answers, and visual breakdowns.

You configure Insights in **Step 3 (Insights)** of the AI Actor setup wizard.

<Frame caption="The Insights Configuration table and the Add Insight panel.">
  <img src="https://mintcdn.com/wittingai/29PVoPfbm1UtpFr0/images/how_to/actor_insights.png?fit=max&auto=format&n=29PVoPfbm1UtpFr0&q=85&s=9654d719a494e2300083f551babb2583" width="1920" height="1440" data-path="images/how_to/actor_insights.png" />
</Frame>

## Auto-generating insights

Click **Auto-generate Insights** to let the AI suggest a starting set of insights based on your Actor's purpose. You can edit or remove any of the generated insights afterward.

## Adding an insight manually

Click **+ Add Insight** to open the side panel. Fill in the following fields:

| Field                          | What it means                                                                                                                                                                            |
| ------------------------------ | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Insight Name**               | A label for the data point (e.g., "Skill Level Assessment").                                                                                                                             |
| **Purpose**                    | A plain-language description that tells the AI what to extract (e.g., "Assess the candidate's skill levels across various UI/UX competencies and visualize the scores on a bar chart."). |
| **Output Type**                | How the data is displayed — Text, Score, Bar Chart, Line Chart, or Pie Chart.                                                                                                            |
| **Factors**                    | The breakdown categories to measure against (e.g., design thinking, user research, interaction design). Only applies to chart output types.                                              |
| **Y Axis (Label)**             | The measurement label for the vertical axis (e.g., "Skill Level (out of 10)"). Only applies to chart output types.                                                                       |
| **Show in insights & reports** | Toggle whether this insight appears in the meeting results dashboard.                                                                                                                    |

## Insight output types

| Type           | Best for                          | Example                                                                      |
| -------------- | --------------------------------- | ---------------------------------------------------------------------------- |
| **Text**       | Open-ended answers, summaries     | "AI Summary", "Salary expectations"                                          |
| **Score**      | Numeric ratings, years, counts    | "Overall Score", "Years of experience"                                       |
| **Bar Chart**  | Comparing categories side by side | "Skill Level Assessment" broken down by design thinking, user research, etc. |
| **Line Chart** | Trends or progression             | "Engagement level over time"                                                 |
| **Pie Chart**  | Distribution or composition       | "Technology stack breakdown"                                                 |

## Insights table columns

Each insight in the configuration table shows:

| Column        | What it means                                                                                 |
| ------------- | --------------------------------------------------------------------------------------------- |
| **Insight**   | The name of the data point.                                                                   |
| **Data Type** | The output type — Text, Score, Bar Chart, Line Chart, or Pie Chart.                           |
| **Factors**   | Tags showing the breakdown categories for the insight (e.g., design thinking, user research). |
| **Include**   | Toggle to include or exclude the insight from reports without deleting it.                    |
| **Actions**   | Menu to view, edit, or delete the insight.                                                    |

## Writing good purpose descriptions

The purpose description tells the AI what to look for. The more specific you are, the more accurate the extraction.

<Tabs>
  <Tab title="Good">
    ```
    Name: Skill Level Assessment
    Purpose: Assess the candidate's skill levels across various
    UI/UX competencies and visualize the scores on a bar chart.
    Factors: design thinking, user research, interaction design,
    visual design, problem-solving
    Y Axis: Skill Level (out of 10)
    ```
  </Tab>

  <Tab title="Vague">
    ```
    Name: Skills
    Purpose: Check skills.
    ```
  </Tab>
</Tabs>

## Example insights

Here is a set of insights for a UI/UX interview AI Actor:

| Insight name                    | Data type | Factors                                |
| ------------------------------- | --------- | -------------------------------------- |
| Overall Summary                 | Text      | candidate strengths, +3                |
| Overall Score                   | Score     | design thinking, user research, +3     |
| Important Question with Excerpt | Text      | important question, candidate response |
| Skill Level Assessment          | Bar Chart | design thinking, user research, +3     |
| Usability Awareness Evaluation  | Pie Chart | usability principles, +3               |

<Tip>
  Start with 4-8 insights. Too few and you miss useful data. Too many and the AI spreads its attention thin. Focus on the data points your team actually uses to make decisions.
</Tip>

## Next steps

<Columns cols={2}>
  <Card title="Test Actor" icon="flask" href="/vocalis/how-to/test-actor">
    Test your insights before going live.
  </Card>

  <Card title="Meeting Results" icon="chart-mixed" href="/vocalis/how-to/meeting-results">
    See how insights appear after a conversation.
  </Card>
</Columns>