Power Platform Community Forum Thread Details

Hi everyone,

I’m building an agent in Copilot Studio and I’m trying to design a solution that can return combined results from Dataverse and SharePoint Word document content.

🔹 Background

I have a Power Platform solution with:

Dataverse table storing document metadata (e.g., Title, Region, MINumber, etc.)
SharePoint document library storing the actual Word documents
Each Dataverse record contains the SharePoint URL / location of the corresponding Word document

So the Dataverse record acts as the metadata layer and SharePoint stores the document content.

🔹 Requirement

I want the Copilot agent to:

Answer questions based on Dataverse metadata
AND also perform generative / semantic search on the content of Word documents stored in SharePoint
Then return combined results where both conditions are satisfied

🔹 Example Scenarios

Let’s say: For a Dataverse record title: “Luxembourg YellowBelt Review”, metadata is Region = “EU” and MINumber = “657556”

The Word document stored in SharePoint contains a clause called “Reservation of Rights”.

I want the agent to correctly answer prompts like:

“Get me all documents which have MINumber = ‘657556’ and have a clause called ‘Reservation of Rights’.”

→ Result should include documents where:

Dataverse.MINumber = 657556

Word document content (SharePoint) contains “Reservation of Rights”

Categories:

Copilot Studio skills development

General topics

Hi @Aveeyn,

Seems like you’ve got a hybrid datasources for a target solution:

Dataverse table holds metadata (Title, Region, MINumber, etc.).
SharePoint library holds the actual Word documents.
Each Dataverse record stores the SharePoint URL of its corresponding document.

The challenge is to return combined results — metadata + document content — in one queryable experience.

Before everything - you get an idea how to build knowledge (with external documents, synonyms and glossary) you can spend some times here, if you are just starting.

In a minimalistic way, we can go for Virtual Tables (Dataverse + SharePoint), let's see if these steps can be achieved.

Create a Dataverse virtual table that surfaces SharePoint document metadata alongside Dataverse records.
Use Power Automate custom connector or Dataverse plug-in to pull document text when queried.
This way, users query Dataverse directly but see combined metadata + document content. So Virtual table is your knowledge source - you can now follow the link I shared.

Or without external services, we can leverage PowerAutomate and AI builder:

Trigger: User searches or requests a record.
Dataverse step: Query the table for metadata.
SharePoint step: Use the stored URL to fetch the Word document.

Use Word Online (Business) connector → “Extract text from Word document.”
Or use AI Builder – Text Extraction to parse .docx.

{
  "Title": "White Board Logistics Pvt Ltd",
  "Region": "South Asia",
  "MINumber": "ABC11345",
  "DocumentContent": "Logistics details and project scope..."
}

4. Compose: Merge metadata + extracted text into a single JSON object.

5. Return: Send this combined object back to Power Apps or Teams. So your data source is unified, you can use this as knowledge source and create prompts what you need on chatbot.

To be very simplistic - if you can bring all the sharepoint documents as knowlege source (if not pile of documents there) no additional flow is needed.

I am sure some clues I tried to give. If these clues help to resolve the issue brought you by here, please don't forget to check the box Does this answer your question? At the same time, I am pretty sure you have liked the response!