Organizations want scalable automation and human empathy. Traditional handoffs via CCaaS replace your bot in the conversation. This guide shows how to keep Copilot Studio in control of Microsoft Teams while enabling bidirectional messaging with your live agent platform—preserving native capabilities, reducing orchestration complexity, and improving customer experience.

Executive Summary

Automation scales support, but human empathy closes complex cases. Traditional CCaaS handoffs replace your bot in the conversation, forcing orchestration through an engagement hub and dropping native channel features. This guide shows how to keep Microsoft Copilot Studio in control of Teams while enabling two-way communication with your live agent platform via an M365 Agents SDK skill. The result: a seamless escalation experience, preserved Teams capabilities, and simpler architecture.

What Is the Copilot Studio Handover Pattern?

When a user requests a human agent, a handover skill (HandoverToLiveAgentSample) creates a session in your customer service system and continues to route messages while Teams stays the active channel. It uses proactive messaging to deliver agent replies back to the user asynchronously.

Message Flow (simplified):

Customer → Teams → Copilot Studio → Skill → Live Chat → Human Agent
Human Agent → Live Chat → Skill → Teams (proactive) → Customer

Users can end the handoff at any time and return to the Copilot agent.

Why This Matters (Benefits)

Preserve Native Teams Features: Mentions, adaptive cards, threading, notifications stay intact.
Simpler Architecture: Avoid double intent recognition and orchestration inside CCaaS virtual agents.
Better Customer Experience: No widget switch; users stay in the same Teams thread.
Vendor-Neutral: Integrate ServiceNow, Salesforce Service Cloud, Genesys, Zendesk, etc.
Operational Control: Keep analytics, governance, and compliance firmly within M365.

How to Implement (Step-by-Step)

Prerequisites

.NET 8 SDK and later
Visual Studio or VS Code
Dev Tunnels CLI for local
Azure subscription
Microsoft 365 tenant (Copilot Studio + Teams)
Dataverse environment
Permissions: Entra ID App Admin or Global Admin; Copilot Studio Environment Maker

Core Components

HandoverToLiveAgentSample (Skill): Handles auth, session lifecycle, proactive Teams messaging.
ContosoLiveChatApp: Demo live chat UI + REST APIs (/start, /send, /end).
HandoverAgentSample.zip: Copilot agent topics: Escalate to Live Chat, Goodbye Live Chat, environment variables.

Local Setup (Dev Tunnel)

Expose your local skill securely for Copilot Studio to reach it:

devtunnel login
devtunnel create  --allow-anonymous
devtunnel port create  -p 5001
devtunnel host

Copy the public URL: https://-5001.euw.devtunnels.ms

App Registrations

Skill App Registration (Entra ID)
- Create app; save Application (client) ID and Client Secret.
Copilot Studio Agent App Registration
- Automatically created on solution import; create Client Secret and note the Agent App ID.

Run the Apps

# Skill
dotnet run --project .\HandoverToLiveAgentSample\HandoverToLiveAgentSample.csproj

# Live Chat demo
dotnet run --project .\ContosoLiveChatApp\ContosoLiveChatApp.csproj

Configure the Copilot Solution

Import HandoverAgentSample.zip to Dataverse.
Set environment variables:
- endpointUrl → https://-5001.euw.devtunnels.ms/api/messages
- msAppId → Skill App Client ID

Configure appsettings.json (Skill)

CopilotStudioBot: TenantId, ClientId (Agent App ID), Secret (Agent Secret).
LiveChat: TenantId, ClientId (Skill App ID), Secret (Skill Secret).

This enables the skill to authenticate as the Copilot agent and send proactive Teams messages back to the user during live sessions.

Update Skill Manifest & Branding

skill-manifest.json → set endpointUrl and msAppId.
Entra → Branding & properties → Home page URL → .../api/messages.
Restart the skill and publish the agent to Microsoft Teams.

Production Checklist

Deploy the skill to Azure App Service (not Dev Tunnel).
Store credentials in Azure Key Vault; rotate regularly.
Replace in-memory mappings with Azure Table, Cosmos DB, or SQL.
Add retry policies, circuit breaker, idempotent webhooks, graceful degradation.
Validate requests (JWT/OAuth), rate limit, enforce TLS, audit access.
Handle PII with retention policies and encryption; log only what’s necessary.
Use Application Insights for traces, metrics, and dashboards.

Customizing for Your Live Chat Platform

Replace the sample integration points inside the skill:

Create Session: EndConversationAction.cs → replace /api/livechat/start.
Forward Message: SendMessageAction.cs → replace /api/livechat/send.
Receive Agent Messages: MessagesController.cs → /api/messages/proactive.
End Session: SendMessageAction.cs → replace /api/livechat/end.

Known Limitations & Risks

Teams-Only Implementation: The pattern is channel-agnostic, but this sample uses Teams proactive messaging.
Concurrent Sessions: Prevent multiple active live chat sessions for one Teams thread.
State Storage: In-memory mappings are not production-safe—use persistent storage.
Platform Auth: Implement proper OAuth/JWT validation for your CCaaS/live chat platform.

FAQs

Q1. Can I use this with ServiceNow, Genesys, or Salesforce?
Yes—replace the sample REST calls with your platform’s chat APIs and webhook endpoints.

Q2. Do users leave Teams during handoff?
No. Copilot Studio keeps control of the Teams conversation, preserving native capabilities.

Q3. Does this support proactive messages?
Yes. Live agent replies are sent asynchronously back to Teams via proactive messaging.

Q4. What’s the long-term Microsoft pattern?
Agents SDK skills work today, but plan for multi-agent orchestration as the recommended forward path.

Copilot Studio Handover to Live Agent: What It Is, How to Implement, Benefits & Limitations