Note
Access to this page requires authorization. You can try signing in or changing directories.
Access to this page requires authorization. You can try changing directories.
[This article is prerelease documentation and is subject to change.]
Use the Customer Conversation harness when your agent needs to answer questions and resolve issues for your customers over voice and chat. An agent built on this harness is called a Customer Assist Agent.
Note
The harness used to build this agent uses usage-based billing. Using, building, testing, and evaluating agents might consume Copilot Credits. Learn more in Overview of usage-based billing.
What is the Customer Conversation harness?
The Customer Conversation harness is best when your agent needs predictable responses for defined topics while still providing a fluid, natural-language conversation over voice or chat. It uses a multi-agent architecture with specialized sub-agents, each with its own escalation path to route unresolved customer questions or issues to a human representative, like in contact centers or customer service. The harness can call tools across connectors, knowledge, Model Context Protocol (MCP), and connected agents, and it supports skills and memory. Use it when you want to publish an agent on a website so customers can chat with it, or assign it a phone number so customers can call it. Agents on this harness use Copilot Credits.
Key concepts
Agents powered by the Customer Conversation harness are organized around their core components. You can configure the following from the Build tab:
Instructions: Define the primary agent's identity, tone, scope, and behavior. Only the primary agent's instructions can include modality-specific guidance, separated into Speech and Chat tabs. Learn more in Configure agent details and instructions (Customer Conversation).
Knowledge: Connected data sources ground the agent's responses in your organization's data. Speech channels don't support end-user sign-in, so knowledge sources that require end-user credentials only work for Chat conversations. Learn more in Add knowledge to an agent (Customer Conversation).
Tools and skills: Tools let the agent perform actions and connect to services, including prebuilt contact-center tools like Escalate (to Teams Phone, email, or a live agent through Customer Engagement) and Biometric Authentication (caller authentication). The Microsoft IQ section lets you connect the agent to Foundry IQ, a managed knowledge layer that gives permission-aware, cited answers from your organization's data. Learn more in Connect to Foundry IQ (Customer Conversation). Skills define reusable structured behaviors, including generative skills and deterministic skills (topics).
Sub-agents and connected agents: Add specialist agents to handle focused intents. A sub-agent shares configuration, deployment, and ALM with its primary agent. A connected agent, such as an agent hosted in Azure AI Foundry, keeps independent configuration and ALM so it can be reused across multiple orchestrations. Learn more in Connected agents overview.
Events: Turn built-in responses for common conversation moments on or off, such as Conversation Start, End of Conversation, and On Error. Learn more in Events overview.
Agent authoring surface
Agents powered by the Customer Conversation harness use a tab-based surface for building and managing:
| Tab | Purpose |
|---|---|
| Build | Configure the primary agent's identity, knowledge, tools, skills, and sub-agents or connected agents |
| Preview | Test the agent interactively in a preview chat interface |
| Evaluate | Create and run test sets to measure agent quality |
| Deploy | Deploy the agent to Speech and Chat channels |
| Monitor | Review recent and live conversation activity |
Agent lifecycle
Building an agent with the Customer Conversation harness follows this general flow:
- Create: Start from the Copilot Studio home page or the Agents tab, and describe your agent in natural language to generate a starting primary agent, or start from a blank agent.
- Build: Configure the primary agent's instructions, knowledge, tools, and skills, then add sub-agents or connect existing agents to handle specialist intents.
- Test: Use the Preview tab to test conversations interactively, and the Evaluate tab to run test sets and review pass/fail results.
- Publish: Use the Deploy tab to publish the agent to Speech and Chat channels.
- Monitor: Use the Monitor tab to track live and recent conversation activity after deployment.