Finance | Project Operations, Human Resources, ...

Access runtime values from the active form from Copilot Chat

(1) Share

Report

Posted on by Diego Mancassola

859

Hi everyone,

I’m currently working with Copilot Chat in Dynamics 365 Finance & Operations and trying to understand whether there’s a reliable way to access runtime values from the active form (e.g., values of controls) directly from a Copilot chat interaction. For example these forms values:

I've read some documentation and i read some "actions", but not clear if it's possible to read/write data of form controls?

On documentation the is that: For example, through client plugins, Copilot in finance and operations apps can be extended to let users enter natural language prompts in Copilot to perform application actions, fill in form values, or ask questions that require calculations and business logic from the application. https://learn.microsoft.com/en-us/dynamics365/fin-ops-core/dev-itpro/copilot/copilot-client-plugins#form-actions

Any one has a similar experience?

Categories:

Dynamics 365 Finance

I have the same question (0)

All responses (4)

Answers (0)

Sort by

Suggested answer

Diego Mancassola 859 on at

Like (0)

Report
Copy link

Link copied!

Hi all, is not well documented on official documentation, but at the end the solution is a x++ class as described the solution provided by Ariste.

https://ariste.info/2025/09/create-client-plugins-copilot-studio-d365-fo/

The key to understand, not so clear for me at begin, is the action is invoked on client side, that allow you to access the runtime formrun object.

Summary, you have to create:

* The class where the execution action will get values from form run instance, you can do whatever you want in that class but in my case i need only callback values from form. Pay attention to returned types, only some is supported. String, Integer, Real, Int64 and Enum but only boolean and NoYesId

* Create a topic where the action is invoked, pass here the input JSON

* Create a secondary topic will handle the call back event, where you get back a JSON with you values

Refer to article to complete solution.

Was this reply helpful? Yes No
Diego Mancassola 859 on at

Like (0)

Report
Copy link

Link copied!
Since the communication between topics is now working correctly and I can pass filters and context from LedgerAdvancedSettlement back to Copilot Chat, the remaining challenge is process orchestration.

My goal is to guide the user through the Ledger Settlement process directly from Copilot Chat:

read filters from the active form (Works)

guide the user step-by-step (Agent)

For do that, i have enabled the generative mode, and add the ERP MCP to standard "Copilot for Finance & Operations Apps" agent.

I tried different approaches:

Child agent: technically starts, and manage correct the parameters, but it enters a loop and does not properly continue the guided workflow

External agent: Already works externally, but not integrated into FO, only on Teams, seems not usable inside standard chat because integrated mode seems unsupported

At the moment I’m not understanding what is the recommended architecture for these guided business processes inside Copilot for Finance & Operations Apps.

Any advice or best practice would be really appreciated.

Was this reply helpful? Yes No
André Arnaud de Cal... 304,686 Super User 2026 Season 1 on at

Like (0)

Report
Copy link

Link copied!

Hi Diego,

It is not clear to me what exactly you want to achieve here. Do you want the sidecar to be able to check if the user is following the correct steps? For what reason do you need to be able to read the filter values and what exactly do you mean by guiding the user?

Was this reply helpful? Yes No
Diego Mancassola 859 on at

Like (0)

Report
Copy link

Link copied!
Not exactly. My goal is more to create a contextual AI workflow directly inside the standard F&O sidecar chat.

The idea is:

the user configures filters in the standard form or in general read values from it

the sidecar reads that runtime context

the agent analyzes the resulting business data through MCP/OData

then guides the user conversationally through a multi-step process to manage business data

So the filter values are important because they represent the runtime context already prepared by the user in the UI. I don’t want the user to re-enter the same information again in chat.

By “guiding the user” I mean an iterative interaction where the agent:

analyzes the current dataset/context

proposes actions or combinations

asks for confirmations/refinements

recalculates results if needed

and eventually executes the final action

Conceptually, I’m trying to use the embedded sidecar not only as a Q&A assistant, but as a contextual process-oriented AI experience directly connected to the active ERP form/session.

The technical challenge I’m hitting is mainly around orchestration continuity and agent integration inside the embedded Copilot experience.

Was this reply helpful? Yes No