Backfill Dataverse prompt columns on existing records

You add a prompt column to a table with 40 000 accounts, and it stays empty on all of them. That is by design: a prompt column runs when a record is created or when one of its input columns changes. Until now, filling the existing rows meant "touching" them with a flow or a bulk update. Dataverse now has a proper option for it, called Apply prompt to existing records in bulk. Here is how to use it without burning your credits.

The problem with the old workaround

Updating every record to trigger the prompt works, but it has side effects: it changes modifiedon and modifiedby, fires every plug-in and flow registered on update, and gives you no estimate of the AI credits you are about to spend. The new option runs the prompt as an asynchronous background job, without pretending to update your business data.

Step 1: allow it in the environment

Two settings control the feature, in the Power Platform admin center under your environment's Settings > Features:

Setting	Default	What it does
Allow bulk processing of existing records	Off	Enables the bulk option in this environment.
Require review before processing imported prompt columns	On	When a solution with a bulk job is imported, the job waits in System jobs until someone starts it.

Leave the second setting on in test and production: it is your safety net against a solution import that silently launches a costly job.

Step 2: run the prompt on existing records

1. In Power Apps, open the table and its prompt column.
2. Check Also generate values for Existing records in bulk.
3. Review the number of records and the estimated credit consumption shown in the panel, then save the column.
4. Select Run prompt to queue the job.

The job runs in System jobs, where you can follow it and see the credits it consumed. In the admin center, System jobs is under your environment's Settings > Audit and logs; its Actions menu lets you cancel, postpone or pause a job (a cancelled job cannot be resumed). A banner on the column panel also shows the status of the latest job.

Step 3: limit what it processes

Prompt columns accept filter conditions: the prompt only runs for records that match them, which is also how you keep a backfill small. The documentation recommends three precautions:

• apply a filter so that the prompt runs only on the records you need (for example, active accounts with a non-empty description);
• test on a few records first, to measure the real credit consumption on your data;
• check the credits available in the environment before you start.

Remember that prompt columns consume Copilot Credits (and AI Builder credits while they still exist): a backfill on a large table is a budget decision, not a technical detail.

Follow the result row by row

Every prompt column comes with two companion columns, <column>_PromptColumnStatus and <column>_PromptColumnDetails. Add them to a view to see which rows succeeded, failed or were skipped:

Status	Meaning
0	Not started
1	In progress
2	Completed
2000	Completed, skipped because the record did not match the filter
2001	Completed, skipped because no input column changed
3	Failed: the details column holds the error

Failures are usually permissions on an input column or missing credits.

Watch out for these

• Jobs expire after seven days. A job is cancelled automatically after seven days, whatever its status. After a solution import, review and start it promptly.
• History is short. A successful job stays visible in System jobs for 30 days, a failed or cancelled one for 60 days.
• Changing the prompt does not recompute existing values. Dataverse has no on-demand execution: a new prompt definition only applies when an input column changes, or when you run a new bulk job.
• Block unmanaged customizations. Creating or editing a prompt column requires this environment setting to be off.
• Availability. As of November 14, 2026, Microsoft Learn says the capability "is in process of rolling out" and may not be available in your region yet. Check the Features page of your environment before planning on it.

Key takeaways

Bulk generation turns prompt columns into a real enrichment tool for existing data, not only for new records. Turn it on per environment, keep the review setting for imported jobs, filter before you run, and test on a handful of rows to know what the full run will cost.

Researched and drafted with AI assistance, then reviewed, edited and published by a human.

Sources

1. Prompt columns in Microsoft Dataverselearn.microsoft.com
2. Monitor system jobslearn.microsoft.com