Power Platform Community Forum Thread Details

Hi everyone,

I'm using a Dataverse Prompt Column in a model-driven app to generate an executive summary for Opportunity Candidate records.

The prompt column works correctly when records are updated individually. However, I'm seeing inconsistent behavior when attempting to trigger regeneration across many records.

Environment

Dataverse Prompt Column
Model-driven app
Opportunity Candidate table
On-demand background workflow that updates the Notes Summary field by appending a space to force prompt re-execution
Prompt column execution enabled

What I'm seeing

When I update a single record:

Workflow completes successfully
Prompt column regenerates successfully
Summary is generated as expected

When I update multiple records:

Workflow jobs complete successfully
Notes Summary is updated successfully
Prompt column frequently returns "AI field generation failed"

What's interesting is that if I later rerun the workflow on one of the failed records individually, the prompt column often generates successfully.

Troubleshooting performed

Confirmed the workflow itself completes successfully.
Confirmed the Notes Summary field is actually changing.
Confirmed the prompt column executes when the field changes.
Tested records with:
Blank Notes Summary
Large Notes Summary
Minimal Notes Summary
Individual reruns typically succeed.
Failures seem much more common when many records are updated within a short period.

Question

Has anyone experienced Prompt Column generation failures when triggering regeneration across many records at once?

Are there known:

Throttling limits?
Concurrency limits?
Queue/capacity limits for Prompt Column execution?
Best practices for bulk regeneration after making prompt changes?

I'm trying to determine whether this is expected behavior for Prompt Columns under load or if there is a recommended approach for refreshing AI-generated values on a large set of existing records.

Categories:

AI Builder

Governance & administering

Microsoft Dataverse with Power Apps

Error handling

Hi,

This sounds like it could be related to capacity, throttling, or concurrency limits behind Prompt Column processing rather than an issue with your workflow. The fact that individual reruns often succeed, while bulk updates result in "AI field generation failed," suggests the prompt generation service may be struggling when many requests are submitted simultaneously.
A few things I'd check:

Monitor Copilot/AI capacity consumption during the bulk refresh.
Reduce the batch size and see if the failure rate decreases.
Add delays or process records in smaller chunks rather than triggering hundreds at once.
Review whether any records have unusually large input content that could increase processing time.

Since the workflow itself completes and the field updates are occurring, the failure appears to be happening during the AI generation phase rather than the Dataverse update. If the behavior is reproducible at specific volumes, it may be worth opening a support ticket to confirm whether there are documented throughput or concurrency limits for Prompt Columns in your environment.

Thanks & Regards,

Manoj Annavajjala