> For the complete documentation index, see [llms.txt](https://docs.askyazi.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.askyazi.com/results-and-reporting/insights.md).

# Insights

See the themes in your AI interview conversations and check the responses behind each one.

The **Insights** tab groups what participants said into themes and answer categories, and links every theme back to the responses behind it. Use it for a first thematic read of your qualitative material, above all AI interview conversations. You find it under **Reporting > Insights**.

{% hint style="info" %}
**In short:**

* Insights reads your study's conversations, mainly AI interviews, and sorts them into question categories and answer categories.
* Every category links to the actual responses, and **View full transcript** opens the whole conversation.
* It is an AI first pass. Check the responses behind a theme before you report it.
  {% endhint %}

<figure><img src="https://358667285-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FzMRQUXiGQlBwWdg6WEmG%2Fuploads%2FVoZsiw1v2D4U0lfqcfKA%2Ffs-insights.webp?alt=media&amp;token=18108814-9412-4754-b883-3fbb80db93b2" alt="Reporting > Insights: a campaign overview, then each theme with its share of answers and the classified responses. Generate insights refreshes it."><figcaption></figcaption></figure>

## What the page shows

* **Study overview** at the top: a written summary of the study's findings.
* **Categories down the left**: the types of question that were asked (for example "Affordability and cost"), each with a count.
* **Answer categories**: for the category you select, the patterns in how people answered, each with a count and share of answers.
* **Classified responses**: the actual question-and-answer pairs behind the category you select. Click **View full transcript** on any response to open that participant's whole conversation in a side panel.

<figure><img src="https://358667285-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FzMRQUXiGQlBwWdg6WEmG%2Fuploads%2FYLNBgbOmV8iHClnZMU5y%2Finsights-categories.webp?alt=media&amp;token=a385a4fb-9dea-4115-b383-d55a65e4ffa3" alt="Insights answer categories with the share of responses in each"><figcaption></figcaption></figure>

## How the analysis works

Yazi analyses the conversations in the background once there are enough responses. It works in four stages:

1. It reads your research brief, questions and a sample of what was asked, and creates question categories that fit your objectives.
2. It puts each AI interview question into one question category.
3. Within each question category, it creates answer categories that describe the patterns in how people replied. These are specific rather than broad.
4. It tags each answer with every answer category that applies. One answer can carry several.

If a study has not been analysed yet, the page tells you so rather than showing a blank screen.

## Manage categories

You can correct the AI's work. Open the **Manage categories** window to review and adjust the categories the AI created.

<figure><img src="https://358667285-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FzMRQUXiGQlBwWdg6WEmG%2Fuploads%2FerwXGJtGNOAN16gxbpmM%2Finsights-manage-categories.webp?alt=media&amp;token=4e014797-d9c4-4855-82e8-cb32f7862f26" alt="The Manage categories window in Insights"><figcaption></figcaption></figure>

## What Insights reads

Insights is built for qualitative material, above all AI interview conversations. For survey questions, use [Graph Data](https://docs.askyazi.com/results-and-reporting/graph-data-and-cross-tabs.md): it charts closed questions and shows themes for open-text and voice note answers.

## Export

Insights can be exported, including as a PowerPoint deck and an Excel file.

## Limits and things to know

* **It needs volume.** On a handful of responses the categories read thin.
* **Short replies can crowd a category.** Answers such as "yes", "ok" or a thumbs-up can be grouped as their own category and take up space. Look past them to the substantive categories.
* **Check before you report.** Themes and counts are an AI first pass. Read a sample of responses behind any theme you put in front of a client.
* **Uploaded enrichment data is never sent to the AI**, so it does not shape the themes.
* **AI must be on** for your organisation.

## Related

* [Reports and PowerPoint](https://docs.askyazi.com/results-and-reporting/reports-and-powerpoint.md)
* [Graph Data and cross-tabs](https://docs.askyazi.com/results-and-reporting/graph-data-and-cross-tabs.md)
* [Transcripts](https://docs.askyazi.com/results-and-reporting/transcripts.md)
* [AI interviews](https://docs.askyazi.com/study-types/ai-interviews.md)
