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

# Build with AI

> Describe what you want. AI builds the dashboard.

Describe what you want to track. The AI builds the dashboard.

The AI Dashboard Builder takes a plain-English description and assembles a dashboard with the right tiles, metrics, and breakdowns. You can tweak anything afterwards.

<Frame caption="AI Dashboard Builder: prompt to finished dashboard">
  <img src="https://mintcdn.com/mentionme/8q3ChbAON-y-ILjT/images/knowledge-v2/reporting/gifs/ai-dashboard-builder.gif?s=c2b3fc39a7ccc692fab685d264633428" alt="AI Dashboard Builder end-to-end" width="876" height="720" data-path="images/knowledge-v2/reporting/gifs/ai-dashboard-builder.gif" />
</Frame>

## When to use it

AI works best for big-picture asks and exploratory questions:

* "Show me how the programme is doing overall vs last month."
* "Where am I losing customers in the funnel?"
* "Which markets are growing fastest?"

It is less good when you already know the exact tiles you want. For that, [build from scratch](/knowledge-v2/referral/reporting/get-started/your-first-dashboard) is quicker.

## How to start

Go to **Referral → Reporting** and click **New dashboard**. The dialog opens with three paths: a prompt textarea, three suggestions, and links to Blank or Browse templates underneath.

You have two ways to kick off AI:

1. **Click a suggestion.** Three starters are ready to go.
2. **Type your own prompt** in the textarea.

### The three suggestion starters

* **Programme overview**. Headline KPIs vs last period, revenue trend, and conversion funnel.
* **Campaign performance**. Revenue, new customers, and share rate by campaign.
* **Funnel analysis**. Conversion from impression to new customer, by market.

## Writing a good prompt

The AI gets better results with a little specificity. Include:

* **Time period**. "For the last 30 days", "year to date", "this quarter vs last quarter".
* **The decision you are making**. "I want to know whether campaign X is outperforming campaign Y."
* **Breakdowns of interest**. "Split by market", "by share type", "per campaign".

A vague prompt ("how are we doing") still works but will give you something generic.

## During the conversation

Once you submit, the dialog switches to chat mode. You'll see the AI stream its reasoning and tile-building steps. The header shows a running count of tiles in the draft.

You can keep chatting to refine:

* "Add a market breakdown to the revenue chart."
* "Swap the time period to year to date."
* "Remove the markdown intro tile."

When you're happy, click **View dashboard**. The dashboard is saved and opens.

## Limitations

* Review the result before saving. The AI sometimes picks the wrong metric or the wrong chart type. Use [choosing a chart](/knowledge-v2/referral/reporting/building-dashboards/choosing-a-chart) as a sanity check.
* Anything you can do by hand, you can still do afterwards. AI is a starting point, not a sealed result.
* The AI can only use metrics and breakdowns that exist in Reporting. If you ask for something we do not measure, it will substitute the closest available.

## Next

* [Choosing a chart](/knowledge-v2/referral/reporting/building-dashboards/choosing-a-chart)
* [Editing the layout](/knowledge-v2/referral/reporting/building-dashboards/editing-layout)