Share via


Service in Microsoft 365 Copilot architecture

This article contains information about the architecture for Service in Microsoft 365 Copilot. The links in the following sections are designed to provide you with in-depth information about security, compliance, data residency/management, AI integration points, and more.

The purpose of these articles and PDF is to help admins, IT professionals, and developers understand how Service in Microsoft 365 Copilot integrates with their Customer Relationship Management(CRM) and the broader Microsoft 365 ecosystem.

Architecture for Service in Microsoft 365 Copilot in a non-Microsoft CRM

The following architecture diagram provides a visual representation of how the Service in Microsoft 365 Copilot agent works in non-Microsoft CRMs. This solution is a generalized architecture pattern, which can be used for many different scenarios and industries.

Screenshot showing the Service architecture in Microsoft 365 Copilot.

Dataflow

All requests are encrypted via HTTPS.

  1. A user in a non-Microsoft CRM desktop submits a prompt for the Service in Microsoft 365 Copilot agent​.
  2. The AI agent requests relevant context and knowledge from the non-Microsoft CRM copilot adapter​.
  3. The non-Microsoft CRM copilot adapter retrieves context and relevant knowledge from the non-Microsoft CRM system​.
  4. The information is provided to the AI agent​​.
  5. The Service in Microsoft 365 Copilot agent​ sends the meta prompt, user prompt, and contextual data to a large language model (LLM).
  6. The Service in Microsoft 365 Copilot agent​ receives the LLM response​.
  7. The Service in Microsoft 365 Copilot agent sends the response back to the user.

Architecture for Service in Microsoft 365 Copilot in Outlook and Teams

PDF document Last update Last updated release
Download Service in Microsoft 365 Copilot architecture document May 21, 2024 May 2024