web
You’re offline. This is a read only version of the page.
close
Skip to main content

Announcements

News and Announcements icon
Community site session details

Community site session details

Session Id :
Power Platform Community / Forums / Power Automate / Power Automate – “Run ...
Power Automate
Suggested Answer

Power Automate – “Run a prompt” (AI Builder) fails in shared flow

(0) ShareShare
ReportReport
Posted on by 98
Hello everyone,
 
I am facing an issue with a Power Automate cloud flow that uses AI Builder – “Run a prompt” and internally accesses a Dataverse AI Model (msdyn_aimodel).
 

Scenario

 

  • The flow works perfectly for me.

  • I shared the flow with a colleague.

  • The colleague:

    • has the same licenses as I do,

    • imported the flow or uses his own connections.


    •  

  • When he runs the flow, the “Run a prompt” action fails.


  •  
 

Error message

 
Action 'Run_a_prompt' failed:
user does not have ReadAccess right(s) for record
of entity AI Model (msdyn_aimodel).

BusinessUnitLevelMinimumPrivilegeDepthRequiredRights = ReadAccess
 
The error references a specific msdyn_aimodel record.

The owner of the AI Model is a different user (not the colleague).
 

What I already tried

 

  • Deleted and re-added the “Run a prompt” action

  • Re-shared the flow

  • Recreated connections for the colleague

  • Verified both users have identical AI Builder / Power Automate licenses

  • The issue only occurs for the colleague, never for me

  • Multiple different msdyn_aimodel IDs fail with the same pattern


  •  
 

Key observation

 

  • The AI Model is a user-owned Dataverse record

  • The colleague seems to have no Read permission on msdyn_aimodel

  • The flow only works for the original creator / owner


  •  
 

My questions

 

  1. Is it expected that AI Builder prompts / AI Models require explicit Dataverse security roles, even with identical licenses?

  2. Is there a way to make an AI Model usable globally or at BU level without:

    • modifying security roles, or

    • forcing the flow to always run under the original owner?


    •  

  3. Is this behavior a known limitation or bug of “Run a prompt”?

  4. What is the recommended best practice for:

    • shared flows

    • multiple users

    • AI Builder prompts?


    •  


  5.  
 
Any insight would be highly appreciated – especially if I’m missing a fundamental design concept here.
 
Thank you very much!
Categories:
I have the same question (0)
  • Suggested answer
    ninihen Profile Picture
    153 on at
    Share the AI Prompt to your colleague then it shall be fine.

  • Suggested answer
    chiaraalina Profile Picture
    2,425 Super User 2026 Season 1 on at
    Hi 
     
    • This is expected behavior, not a bug: AI Builder prompts are user-owned Dataverse records (msdyn_aimodel) with row-level security that requires explicit sharing
    • Identical licenses don't grant access: Dataverse security roles and record-level permissions control access independently of AI Builder licensing
    • Sharing the flow ≠ sharing the model: These are separate Dataverse resources requiring independent sharing operations
    • The fix is straightforward; As already mentioned in the answer before, use AI Builder's built-in Share function to grant colleagues "User" (run-only) or "Co-owner" (edit) access to the prompt
     
     
    Best Pracitces:
     
    • Explicit Model Sharing: Always share AI Builder models with any user (maker or end-user) who needs to run or build flows/apps that call them. Use teams to simplify sharing for larger audiences (share once to a group/BU team rather than each person). Grant User access when colleagues only need to run the model (not edit) and Co-owner access for co-developers who need to modify or retrain the model.
    • Use a Dedicated Service Account or Service Principal (for flows): For critical, shared flows that many people trigger, consider running them under a non-human account with the necessary privileges. Microsoft recommends using a service principal to own such flows in enterprise scenarios, insulating them from individual user departures. This approach centralizes model access (you’d share the model once with the service account) and improves stability, at the cost of some initial setup and licensing.

Under review

Thank you for your reply! To ensure a great experience for everyone, your content is awaiting approval by our Community Managers. Please check back later.

Helpful resources

Quick Links

Season of Sharing Community Challenge Launch!

Jump in, show your community spirit, and win prizes!

Kudos to our 2025 Community Spotlight Honorees

Expanding mentorship, skilling, and AI innovation

Congratulations to the May Top 10 Community Leaders!

These are the community rock stars!

Leaderboard > Power Automate

#1
Valantis Profile Picture

Valantis 377

#2
11manish Profile Picture

11manish 279

#3
David_MA Profile Picture

David_MA 234 Super User 2026 Season 1

Last 30 days Overall leaderboard