Create a full dashboard from a single prompt with Luzmo's Dashboard Agent — available both in Studio and via the API. Instead of building one chart at a time, describe the dashboard you want and let the agent assemble the charts, layout, and filters together.
In this article, you will learn how to:
Generating one chart only requires the AI to pick a visualization type, a metric, and a dimension. Generating a dashboard requires it to also decide:
Because of this, a dashboard-level prompt reads more like a brief than a chart-level prompt. Instead of describing one visualization, describe the outcome the dashboard should support.
In Studio and the Embedded Dashboard Editor (EDE), look for the Agent button and panel in the side menu. The dashboard empty state also points to it as a way to start a new dashboard.
In EDE, the Dashboard Agent is only shown in clients using the updated EDE with flag_embedded_studio_update. It's enabled by default; to disable it, add !flag_dashboard_agent to the feature_overrides array in your Embed token request.
Build me a dashboard tracking weekly active users, retention, and churn by customer segment, for the last two quarters.
A reasonable AI-generated result includes a small set of charts covering each named metric, a segment filter that applies across all of them, and a sensible default layout. From there, you can continue the conversation:
Add a filter for plan tier as well, and make the churn chart a trend line instead of a bar chart.
This mirrors how you'd iterate with a colleague building the dashboard with you — refining scope and layout incrementally rather than starting over.
The prompt box isn't limited to typed descriptions:
If you already know which dataset the dashboard should use, say so up front to keep the result grounded in the right data:
Using the product usage dataset, build a dashboard that shows adoption of our three newest features over the last quarter, broken down by account tier.
Being specific about the dataset (and, where relevant, which columns matter) reduces the chance the AI reaches for the wrong metric or an unrelated dataset. See Preparing your data for AI for more on getting your datasets ready for this.
The Dashboard Agent isn't only for building from scratch — point it at an existing dashboard and it can:
Every change the agent makes can be safely undone with the Undo button below its latest message, so it's safe to experiment with a prompt before committing to it.
Treat the first generated result as a strong starting layout, not a final answer:
Beyond Studio, dashboards can be generated programmatically through the /aiprompt endpoint (agent: 'dashboard'), then persisted with create('securable'). This is useful when you need more than one hand-built dashboard, for example to:
See Generate a dashboard with the Dashboard Agent in the developer docs for the API reference and examples.
You don't need to choose one approach for the whole dashboard. It's common to generate an initial set of charts with AI, then add one or two manually built charts for a specific metric with unusual requirements (a custom formula, a specific formatting need, or a chart type the AI didn't select). AI-assisted and manually built charts behave the same way once they're on the canvas — there's no distinction in how they're edited, filtered, or shared afterward.
Dashboard-level AI assistance is about building the dashboard faster. It's distinct from:
If your goal is a reusable asset other people will come back to and explore, build it here. If your goal is answering a one-off question, IQ is usually the faster path.