Real-Time Dashboards include built-in visuals for common monitoring and analysis scenarios. However, operational decisions do not always fit a standard chart. You may need to combine several signals, use a domain-specific layout, or present information differently for a particular workflow or audience. Custom visuals with Copilot, now in preview, help you create a dashboard experience around the operational question you need to answer.
With custom visuals, you can describe the operational view you need in natural language. Copilot creates the supporting KQL query and HTML-based visualization directly in the dashboard, helping you move from an operational question to a tailored visual without first adapting the scenario to a predefined chart.
Move beyond predefined chart types
A standard chart starts with a visual type such as a line, bar, or pie chart. You then shape the query and formatting to fit that structure. This model works well when the question maps naturally to an established chart.
Operational questions don't always fit that model. You might need a visual that combines several signals, uses a domain-specific layout, or presents status in a form that matches how your team works. Building that experience outside the dashboard can require front-end development and separate deployment and maintenance processes. Custom visuals begin with the outcome instead. Describe the operational question and the visual you want, and Copilot uses the selected database context to generate both the KQL query and an HTML-based visualization. The result is added directly to the dashboard as a new tile.
Example: Turn venue metrics into an operational view
Consider a venue operations dashboard that tracks wait times and ingress volume across several entrances. Standard KPI and chart visuals can show those measures individually, but an operator may need to understand how conditions differ across the venue.
With a custom visual, the dashboard author can describe that operational view directly: arrange entrances spatially around the venue, show the current wait time and ingress volume for each entrance, and use color to highlight congestion. Copilot can then generate a visual design around that scenario.
The result is not simply another way to plot the same data. It presents the information in a form designed around the operational question: Where is congestion building, and which entrance needs attention?
Create a custom visual
Start from a Real-Time Dashboard that you can edit, and that has its data sources to create a custom visual, open a Real-Time Dashboard that you can edit and confirm that its data sources are configured for custom visuals. In Editing mode, select Add visual > Describe a visual, enter your prompt, and submit it. Copilot adds the visual to the current dashboard page and generates the connected KQL query and HTML-based visualization
Figure: In Editing mode, select Add visual, and then select Describe a visual to begin creating a custom visual.Figure: Copilot displays a generated KQL query and the connected custom visual in the dashboard.
Review the first result as a starting point. The quality of the visual depends in part on how clearly the prompt communicates the question, the data to use, and the intended presentation.
Figure: Review the generated entrance operations visual.Write a focused prompt
A useful prompt gives Copilot enough context to choose the relevant data and design a visual that answers a specific question. Include the following details:
- The operational question: Explain the decision or condition that the visual should support.
- Measures: Name the values to calculate or compare, such as event count, latency, volume, or service state.
- Dimensions: Identify how to group or organize the results, such as location, service, category, or severity.
- Time range: Define the period that matters, such as the last hour or the last 24 hours.
- Presentation: Describe the visual form when you have a specific operational layout in mind.
- Interactions: Describe how viewers should explore the visual, such as displaying additional details or using controls to change what is shown.
For example, you can ask Copilot to:
- Show the current operational status of each venue entrance. Display average wait time and ingress volume for each entrance and use color to make congestion immediately visible. Arrange the entrances spatially around the venue so operators can quickly identify where attention is needed.
- Create a storm monitoring view that positions storm events by latitude and longitude. Size each event based on severity and use color to distinguish warning levels. Show additional event details when a viewer explores an event.
- Create a service health board that shows each service and its dependencies as connected nodes. Use color to distinguish healthy, degraded, and critical services, and make critical services visually prominent so operators can quickly identify where investigation is needed.
These prompts do more than name a chart. They describe the relationship between the data and the way a viewer should interpret it. You can make them more precise by adding the relevant measures, status rules, categories, or time window from your data.
Review and refine the result
Creating a custom visual is an iterative authoring experience. After Copilot creates the tile, review whether the visual answers the intended question and whether the generated query selects and aggregates the expected data.
To make changes, select the Edit icon on the tile. In tile edit mode, you can review the query and the visual, then use the Copilot pane to refine either one. Follow-up prompts can clarify the original request without requiring you to start over. For example, you might narrow the time range, change the grouping, emphasize a service state, or adjust how the information is presented.
Continue until the visual communicates the operational state clearly. Then select Apply, resize, or reposition the tile, and save the dashboard. Switch to viewing mode to confirm that the saved visual renders as expected in the experience your dashboard viewers use.
This review step is important for AI-assisted authoring. Validate the generated query and visuals against your data and operational requirements. Clear labels, appropriate aggregation, and a focused visual hierarchy help viewers understand the result quickly.
Next steps
Start with an operational question that does not fit a standard chart. Describe the experience you need and use Copilot to create a first version that you can review and refine before sharing the dashboard. This approach helps you turn specialized monitoring needs into tailored operational views without building and maintaining a separate front-end experience.
Learn how to configure the required identity and cloud connection, create a tile, write a focused prompt, and refine the result in Create a custom visual in a Real-Time Dashboard.
Explore the built-in options in the Real-Time Dashboard visual gallery.
Review formatting options for standard tiles in Customize Real-Time Dashboard visuals.
Microsoft Employee