Power Platform Community Forum Thread Details

"Compare Files with AI"

The feature works well interactively from the SharePoint document library and returns an AI-generated comparison.

What I am trying to determine is whether this capability is available programmatically.

My target architecture is:

User uploads two documents
→ Copilot Studio agent
→ SharePoint
→ Invoke SharePoint AI Compare Files
→ Receive comparison result
→ Agent formats and presents the result

Questions:

Is the SharePoint "Compare Files with AI" capability exposed through:
- Microsoft Graph API?
- Power Automate action?
- SharePoint REST API?
- Copilot Studio tool/action?
- Any other backend endpoint?
Has anyone successfully invoked this comparison capability from:
- Copilot Studio
- Power Automate
- Azure Function
- Custom application
If this is currently a UI-only Copilot feature, is there any published roadmap for backend/API access?
If backend access is not available, what architecture are people using to replicate the same comparison capability for uploaded SharePoint documents?

I would appreciate any documentation, implementation examples, or confirmation from anyone who has successfully automated this feature.

Thanks.

Categories:

Calling actions from Copilot Studio

Hello JS-02060508-0

"Compare Files with AI" is a UI-only Copilot feature in SharePoint. It is not exposed through any public Graph API, SharePoint REST endpoint, Power Automate action, or Copilot Studio tool, and it is not on the public roadmap.

Available through Graph / SharePoint REST / Power Automate / Copilot Studio / any backend? No on all five.
Has anyone successfully called it from Copilot Studio / Power Automate / Azure Function / custom app? No — there is no endpoint to call.
Is there a published roadmap for API access? None. File on the M365 feedback portal if you want it tracked.

Replica architecture (what teams actually ship):

The agent receives the two file URLs/uploads.
It fetches both through Graph: GET /sites/{siteId}/drive/items/{itemId}/content.
Extract clean text — Azure AI Document Intelligence (read/layout model) for PDFs; native parsers for DOCX/XLSX/PPTX.
Run a structural diff first (e.g., redlines lib or git-diff on extracted markdown) to produce deterministic add/delete/modify hunks.
Send the hunks to an LLM (Azure OpenAI or AI Builder Prompt) and ask it only to explain the semantic significance — that cuts hallucination and token cost compared with asking the LLM to find diffs from scratch.
Return structured JSON: { summary, additions[], deletions[], modifications[], semantic_differences[] }.
Display it in the agent as an Adaptive Card / markdown table.

If you found the information above helpful, I would appreciate it if you could share your feedback.|
Your feedback is important to us. Please rate us:
🤩 Excellent 🙂 Good 😐 Average 🙁 Needs Improvement 😠 Poor