Power Platform Community Forum Thread Details

Hi Community,

I'm looking for guidance on the recommended deployment model for a Power Automate Desktop (PAD) solution in a multi-user environment.

Current Solution

Power Automate Desktop (PAD) is used as the RPA layer.
PAD launches a Python executable.
The Python application displays a desktop UI where users provide inputs.
The solution is hosted on an AWS Windows Server.
Around 20 users will access the server through separate RDP sessions.

Options We Are Evaluating

Option 1

Deploy the desktop flow under a shared service account on the AWS server.

Option 2

Register the AWS server as a machine in a Production environment.
Maintain a single desktop flow.
Allow multiple users to trigger the same flow.

Questions

What is the recommended deployment model for this type of solution?
Can a single desktop flow be shared among multiple users, or should each user have their own copy?
If multiple users trigger the same desktop flow simultaneously from separate RDP sessions on the same Windows Server, how does PAD determine which user session to execute in?
Is concurrent execution across multiple RDP sessions supported on a single registered machine?
Has anyone implemented a similar architecture in production? If so, what approach worked best?

I'm primarily looking for guidance on the deployment architecture and execution model rather than licensing.

Any recommendations or best practices would be greatly appreciated.

Thank you!

Categories:

Power Automate Desktop

Hi,

For this type of PAD/RPA solution, I would be careful with the multi-user RDP model.

Power Automate Desktop is generally best treated as a machine/session-based automation, not as a shared desktop application runtime for many users at the same time.

For your questions:

Recommended model

I would recommend registering the AWS Windows Server as a machine or machine group in the target Power Platform environment and running the desktop flow through a controlled account/session.

For production, avoid users manually sharing and running separate copies unless there is a strong reason.

Single desktop flow vs multiple copies

A single centrally managed desktop flow is better. Multiple copies become difficult to maintain, troubleshoot, and version.

You can pass different user inputs into the same flow instead of duplicating the flow per user.

Which RDP session does PAD use?

This is the key concern. PAD does not simply choose any active RDP session from 20 users. Desktop flows execute in the Windows session associated with the machine connection/run configuration. If multiple users are logged into the same server, that can create unpredictable or unsupported behavior for UI automation.

Concurrent execution on one registered machine

Do not design this assuming one registered Windows Server can run many simultaneous PAD UI automations across separate RDP sessions. For reliable production use, assume one interactive desktop flow run per machine/session at a time.

If you need concurrency, use multiple machines/VMs and place them in a machine group so Power Automate can distribute runs.

Best practice

A better production pattern would be:

Cloud flow / app collects user input

↓

Queue the request

↓

Desktop flow runs on a registered machine

↓

PAD launches Python executable

↓

Python processes one request

↓

Result is written back to Dataverse / SharePoint / SQL / email

If 20 users may submit requests at the same time, queue the work instead of trying to run 20 desktop sessions on the same server.

In short, I would use one managed desktop flow, one or more registered machines, and a queue-based approach. If concurrent processing is required, scale horizontally with additional VMs rather than relying on multiple RDP sessions on one Windows Server.