Work with Prompt columns and row summaries

  • 8 minutes

A Prompt column is an AI-powered column type in Dataverse that stores the result of a natural language prompt run against other column values in the same row. When a record is created or updated, Dataverse evaluates the prompt and stores the generated output directly in the column — ready to be used in apps, flows, or views without any manual entry.

This unit covers how to create and configure Prompt columns, and how to configure a related feature — row summaries — that uses the same prompt-based approach to surface key record information directly in model-driven app forms.

Prompt columns

Prerequisites

  • Prompt columns use the same licensing model as AI Builder prompts. For details, see Prompts overview.
  • Copilot and AI Prompts features must be turned on in the environment's Feature settings. For details, see Manage feature settings.

Create a Prompt column

  1. In Power Apps, select Tables in the left navigation and open the table where you want to add the column.

  2. Select New > Column.

  3. Enter a Display name and optional Description for the column.

  4. Under Data type, select Prompt.

  5. Clear the Allow form fill assistance checkbox.

  6. Select + Add new prompt. You can create up to five Prompt columns per table.

  7. On the prompt page, write or customize your prompt.

  8. Select Save on the prompt page, then Save on the column properties page.

Write effective prompts

Prompt columns use natural language instructions to tell the AI model what to generate. Keep these tips in mind:

  • Be specific — Instead of "Summarize the record," write "Summarize the customer feedback in two sentences, focusing on the main issue and the customer's sentiment."
  • Use structured language — "Generate a list of the top three recommended actions based on the service request description" is more effective than "What should we do?"
  • Add input columns — Reference other columns in the table as inputs so the AI has the right data to work with.

Add input columns to a prompt

  1. In the prompt editor, select + Add content and then select the table.

  2. Select the input column from the column dropdown and select Add.

  3. You can add more than one input column per prompt.

Note

Input columns can't be Formula columns, File columns, Image columns, or other Prompt columns — their values are ignored if selected.

Test and refine a prompt

Create a test record with values in all input columns before testing.

  1. In the prompt editor, select the input column reference to open the Filter knowledge pane.

  2. Select Filter attribute, choose the filter column (such as Name), and enter the value of your test record.

  3. Select Close, then select Test and review the Model response.

  4. Check the Knowledge used tab to confirm the response was generated from your test record.

  5. Adjust the prompt until the output meets your needs, then select No filter in the Filter knowledge pane and select Save.

View Prompt column results

Add the Prompt column to a model-driven app form or view to see the generated values. When a record is saved, Dataverse runs the prompt asynchronously and stores the result in the column. Two additional system columns are created automatically for each Prompt column — (columnName)_PromptColumnStatus and (columnName)_PromptColumnDetails — which record the execution status and any error details.

Add filter conditions (optional)

Filter conditions let you control when a prompt runs, so it only executes for records that meet specific criteria. This reduces unnecessary AI credit consumption.

  1. In the prompt editor, open a table variable and select Apply filter.

  2. On the Filter conditions pane, select Add filter and configure the condition, then select OK.

  3. Select Save on the column properties page.

Row summaries

A row summary is a Copilot-generated overview of a record that appears in a collapsible bar at the top of a model-driven app main form, or directly from views. Unlike a Prompt column, a row summary isn't stored as a column value — it's generated on demand when a user views the record.

Use a row summary to help users quickly orient themselves to the most important information in a record without scrolling through all its fields and related tables.

Prerequisites

A Power Platform administrator must have the AI insight cards environment setting enabled. For details, see Manage feature settings.

Note

Some Dynamics 365 app tables (Case, Lead, Opportunity) provide their own built-in summaries via Dynamics 365 Customer Service and Sales. The row summary feature isn't available for those tables.

Configure a row summary

Row summaries are configured at the table level and apply to all main forms and views for that table.

  1. In Power Apps, select Tables in the left navigation and open the table.

  2. Under Customizations, select Row summary. If the option is disabled, hover over it to see the reason (for example, the table must have at least one row of data).

  3. In the Prompt box, define the summary by selecting Add or typing / to reference columns. You can also specify formatting (bulleted list, paragraph) and language instructions.

  4. Select Test prompt to preview the output using the most recently edited record.

  5. When satisfied, select Apply to main forms.

Write a good row summary prompt

  • List the specific columns you want included (or excluded).
  • Specify formatting — for example, "Output as a bulleted list with bold labels."
  • If your users work in multiple languages, include the LanguageCode input and add the instruction: "You must respond in language <LanguageCode>."

Add a row summary to a solution

Row summaries are solution-aware. To move a summary between environments:

  1. In Power Apps, open the solution.

  2. Select Add existing > More > Other > AI Skill Config.

  3. Select the row corresponding to the table's summary and select Add.

For more information, see Configure a row summary for a model-driven app.