Configure agent details and instructions (Customer Conversation) (preview)

[This article is prerelease documentation and is subject to change.]

In the agent details and instructions, define the agent and its behavior. This definition shapes the agent's tone, scope, and response style, and helps the orchestrator route conversations correctly across a multi-agent orchestration.

Agents powered by the Customer Conversation harness depend on instructions to decide what resources (tools, knowledge, skills, or sub-agents) to use to address a user query, and to generate a response to the end user. Only the primary agent's instructions can include modality-specific guidance, separated into Speech and Chat.

Note

The agent can only invoke tools, topics, or sub-agents that you've added and configured on it - it won't attempt to call resources that aren't configured. Instructions can also guide how the agent fills a tool's inputs when it invokes that tool, so describe any formatting or value requirements the tool expects.

Configure the following elements from the Build tab:

Name: The agent's display name.
Description: Up to 1,024 characters. A description of what the agent handles, so the orchestrator knows when to route conversations to it.
Instructions: Natural language guidance that shapes the agent's behavior, tone, and scope. The primary agent has a General tab for instructions that apply across all channels, plus separate Speech and Chat tabs that layer extra, modality-specific guidance on top of the general instructions. Each of these tabs has a hard limit of 20,000 characters. Each sub-agent's general instructions also have a hard limit of 20,000 characters. Learn more in Sub-agent fields.

Configure your agent's details and instructions

Open your agent in Copilot Studio.
Select the Build tab.
Select the primary agent's card to open its configuration panel. If you've added any sub-agents, select a sub-agent's card instead to configure that sub-agent.
In the Name field, enter or update the agent's display name.
In the Description field, describe what the agent handles, so the orchestrator knows when to route conversations to it. This field has a 1,024-character limit. Select See an example for a sample description.
In the Instructions section, on the General tab, describe what the agent does, its tone, and its rules. This field has a hard limit of 20,000 characters for both the primary agent and each sub-agent. Type / to insert tools, topics, and variables. Select See an example for a sample instruction.
For the primary agent, select the Speech or Chat tab located below Instructions to add modality-specific guidance. Each tab has a hard limit of 20,000 characters and layers extra guidance on top of the general instructions:
- Select Speech to guide the agent's speaking style for phone calls. Consider tone of voice, how to handle pauses, phrases to use or avoid, and how formal to sound. For example: "Speak in a warm, professional tone. Keep responses short - no more than 2-3 sentences. Avoid reading out long lists."
- Select Chat to guide the agent's writing style for web chat. Consider response length, formatting preferences, how formal to sound, and whether to use lists or plain sentences. For example: "Be concise and friendly. Use numbered steps for multi-part answers. Avoid jargon."
Select Save.

Write effective instructions

Instructions are the most important tool for shaping agent behavior. Well-written instructions help the agent stay on topic, respond in the right tone, and handle edge cases appropriately.

Consider including:

The agent's primary role and purpose.
The tone and style the agent should use (formal, friendly, concise).
Subjects or tasks the agent should decline or redirect.
How the agent should handle unclear or ambiguous user input.
Any escalation or handoff triggers.

Tip

Start with a clear, simple description of the agent's purpose and expand from there. You can update instructions iteratively as you test the agent's behavior.

Keep instructions concise to preserve conversation context

Agents use both instructions and conversation history as context when responding. The context window is the amount of information an agent can use at one time. Longer instructions fill the context window faster, leaving less room for conversation history.

To help preserve conversation context:

Keep skill instructions to no more than 3,000 characters per skill.
Use no more than 10 skills per sub-agent.
Avoid repeating instructions across the primary agent's Speech and Chat tabs. Put instructions that apply to both modalities on the General tab, and use the modality-specific tabs only for guidance unique to Speech or Chat.

The skill length and skill count recommendations are best practices, not hard limits.

Best practices for instructions across sub-agents

The Customer Conversation harness works best when sub-agents are independent, almost like different departments in a contact center. Independent sub-agents help route the right context to the right escalation paths, such as different contact center queues.

The primary agent's description and instructions shouldn't be mistaken for any of its sub-agents'. Give each agent a distinct role so the orchestrator can tell them apart.
Sub-agents should generally be mutually exclusive, each with a clearly defined role. Give a sub-agent its own set of tools and knowledge so it can answer a particular kind of question well on its own.
To get sub-agents to act more fluidly together, use / in an agent's instructions to reference other sub-agents by name. For example, in the primary agent's instructions: "If a user wants to pay their credit card using a checking account from the same bank, ask the /CheckingAccountAgent to first confirm the balance, then have the /CreditCardAgent make the payment."
The primary agent routes between sub-agents. Think of it as the brains of the routing.

The following table shows what each type of agent can reference:

Referencing	Tools	Topics	Events	Sub-agents	Variables
Primary agent	Yes	Yes	Yes	Yes	Yes
Sub-agent	Yes	Yes	Yes	No	Yes
Connected agent	No	No	No	No	No

Tip

To best preserve context discussed within a topic when control returns to the generative orchestration, save the topic's output as a variable.

Next steps

Feedback

Was this page helpful?

Last updated on 2026-10-22