
Instead of adding widgets one by one, you describe in your own words what you want to see. Indicate reads your description, picks the matching metrics from your connected sources, and builds a complete dashboard from them.
The AI does not invent numbers. It only assembles widgets from the metrics your connected data sources already provide, which is why the result is as good as the sources you have connected.
On top of the widgets, the AI writes a summary of what the data currently shows, so you get the interpretation together with the charts.
You want a first version of a dashboard in a minute instead of building ten widgets by hand.
You know what you want to see, but not which metric in which source delivers it.
You want a written interpretation of your numbers, not just charts.
AI can make mistakes. The summary is generated by AI and is provided for informational purposes only. It may be incomplete or inaccurate and does not constitute professional, financial, or business advice. Always check the underlying figures before making decisions based on it. Independently confirming accuracy.
You start the same way as with any dashboard, but choose the AI route instead of the empty canvas.
In the left sidebar, click New dashboard.
At the top, keep Describe it selected. The alternative, Start blank, gives you the empty canvas.
Click the Dashboard description field and describe what you want to see. Write it in your own words, in full sentences or as a list.
Under Sources, click Choose sources and select the Insights the dashboard should use, for example Matomo.
Click Generate dashboard.
The dashboard opens with the finished widgets. A green banner tells you how many were created, for example Generated 9 widgets from your prompt. Indicate also writes a title and a short description for the dashboard, both based on your prompt.
Note: The counter under the field shows how much of the available prompt budget you use, for example
413 / 8000 (incl. sources). The sources you pick count towards it, so very long prompts with many sources can reach the limit.
The prompt decides what you get, so name the figures you want and how you want them broken down. The panel itself suggests three examples, and they show the pattern well:
Marketing overview: sessions trend, top countries, paid spend vs sessions
Weekly sales: revenue by week, orders by product category, average order value
Ad performance: spend and clicks per campaign, cost per click over time
What works:
Name a topic and then the figures. "Website analytics: visitors, actions, bounce rate, visitors by device."
Say how to break it down. "by country", "by week", "per campaign" turns a single number into a chart.
Ask for comparisons. "paid spend vs sessions" puts two figures next to each other.
What does not work:
Asking for data that none of your sources deliver. The AI can only use metrics that exist in your connected sources.
Very vague prompts such as "show me everything". You get a generic board that you then have to rework.
Picking the sources is optional but marked Recommended, and it changes the quality of the result. It lets the AI draw from the right data products, which cuts the amount of prompt budget used and improves the result. You still describe what you want in the prompt.
With sources picked: the AI works with the metrics of those data products only. Faster, cheaper, and closer to what you meant.
Without sources picked: the AI has to consider everything connected in your space, which uses more of the budget and can pull in metrics you did not have in mind.
Pick the source that owns your topic. For a website dashboard that is your web analytics source, for an advertising dashboard your ad platform.
Next to the widgets, the dashboard carries a written summary of the current data. It is the part that answers "and what does that mean", which charts alone do not.
The summary reads the same numbers the widgets show and describes them in sections: what the key figures did in the period, what stands out, what to check, and notes about the data itself, for example when a widget covers a different time range than the rest of the board.
The summary shows when it was written, for example Generated just now.
Click Regenerate to write it again against the current data.
Note: The summary is a reading of your numbers at that moment. After the next data load, or after you change widgets, regenerate it so the text matches what the dashboard now shows.
You do not have to edit widget by widget. The same prompt mechanism can rework an existing dashboard.
Open the dashboard and click the AI icon in the toolbar at the top right.
In the field, describe the change. The examples in the field show the range: add revenue by country, drop the sessions chart, make the top row bigger.
Optionally pick the sources again under Sources.
Click Apply, or press Ctrl + Enter or ⌘ + Enter.
Note: A change prompt rebuilds the whole board around your change, so widgets you built by hand can be redesigned in the process. If you want to keep a widget exactly as it is, duplicate the dashboard first.
I can no longer create new dashboards with AI.
You have run out of AI credits. Go to Settings → Usage and check whether you still have credits left.
If your AI budget is set too low, raise it there.
If you need more AI tokens, contact support through our request form.
The dashboard is missing a figure I asked for.
The metric probably does not exist in the sources you picked. Check the source article for that integration to see which metrics it delivers, then either pick another source or add the widget by hand.
Widgets show no data.
The AI built the widget correctly, but the source has no values for that period. Set the time range to a longer window and check whether data appears.
The result is not what I meant.
Make the prompt more specific and generate again. Name the figures and the breakdown you want instead of the topic alone, and pick the source that owns the topic.
My hand built widget was changed after a prompt.
That is expected. A change prompt rebuilds the whole board. Work on a duplicate when you want to protect manual work.
The summary does not match the widgets.
The summary was written at an earlier point in time. Click Regenerate to write it against the current data.