Copilot in Dynamics 365 Customer Service architecture

The following architecture diagram provides a visual representation of how Microsoft Copilot for Customer Service works. This solution is a generalized architecture pattern, which can be used for many different scenarios and industries.

Architecture

Dataflow

All requests are encrypted through HTTPS, which is illustrated with the icon in the diagram.

1. A user in Dynamics 365 Customer Service submits a prompt that is sent to Copilot
2. Copilot accesses Dataverse and Semantic Index for preprocessing
3. Copilot sends meta prompt, user prompt, and contextual data to Large Language Model (LLM)
4. Copilot receives LLM response
5. Copilot accesses Dataverse and Semantic Index for post-processing
6. Copilot sends the response

Components

• Dynamics 365 Customer Service
• Manage Copilot features in Customer Service
• Microsoft Dataverse
• Azure OpenAI service

Cost optimization

Cost optimization is about looking at ways to reduce unnecessary expenses and improve operational efficiencies. Learn more at Overview of the cost optimization pillar.

Next step

• Dynamics 365 Customer Service implementation optimization resources

Related resources

Review the following related architecture guides, solutions, and other guidance content:

• Artificial intelligence (AI) - Architectural overview