> ## Documentation Index
> Fetch the complete documentation index at: https://docs.risponse.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Respond and escalate to humans

> Let AI answer routine WhatsApp questions and hand harder issues to your team with private notes and tickets.

The AI agent can answer simple questions on its own. When it cannot help, it brings in a person from your team. This page explains how that handoff works.

## How it works

The agent replies when a customer sends a question, complaint, or request and is waiting on you. If the answer is in your [knowledge base](https://docs.risponse.com/ai/agent/knowledge-base.md), it uses it. If not, or if your rules say a person should step in, it escalates.

It can escalate in two ways:

* **Private note.** An internal message that mentions the right team member. The customer never sees it.
* **Ticket.** A tracked issue assigned to someone. The customer is told a ticket was raised.

The agent is meant to avoid opening duplicate tickets for the same problem in one chat.

## Set it up

**1. Write your instructions**

In the agent's personalization settings, describe your team and when to hand off. For example: "Billing questions go to Priya. Anything about a damaged item becomes a ticket."

**2. Set when the agent replies**

Add any rules for when it should answer or stay quiet. See [Activation](https://docs.risponse.com/ai/agent/activation.md).

**3. Turn on the tools**

Enable private notes and tickets in the built-in tools. See [Built-in tools](https://docs.risponse.com/ai/agent/built-in-tools.md).

**4. Add tagging and assignment rules**

Say who gets tagged or assigned for each type of issue.

**5. Save and test**

Send a few test messages from another phone and check what the agent does.

## Good rules to write

* Name the person or team for each topic.
* List topics the agent must never handle, like refunds or legal threats.
* Say what the agent should tell the customer while they wait.

> [!WARNING]
>
> The agent follows your rules as written, but it can still misjudge a message. Review its notes and tickets often, and check important replies.

## Related

* [AI agent overview](https://docs.risponse.com/ai/agent/overview.md)
* [Create a ticket](https://docs.risponse.com/tickets-and-tasks/create-ticket.md)
* [Flag messages with AI](https://docs.risponse.com/ai/flag-messages.md)


## Related topics

- [Activation](https://docs.risponse.com/ai/agent/activation.md)
- [Built-in tools](https://docs.risponse.com/ai/agent/built-in-tools.md)
- [Automated escalation](https://docs.risponse.com/automations/escalation.md)
- [Workspace metrics](https://docs.risponse.com/features/workspace-metrics.md)
- [AI agent overview](https://docs.risponse.com/ai/agent/overview.md)


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.