> For the complete documentation index, see [llms.txt](https://cortado-ai.gitbook.io/cortado-ai-docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://cortado-ai.gitbook.io/cortado-ai-docs/ai-101/route-to-host.md).

# Route to Host

## Overview

When your AI is set to [**Autopilot Mode**](broken://pages/JwAU7bkBKJhhRd0rkmXW), it will automatically attempt to answer any incoming guest message using the information provided in its sources (see [**What does my AI know?**](/cortado-ai-docs/getting-started/understand-your-ai/what-does-my-ai-know.md) for a comprehensive overview).

However, if your AI lacks the information to fully answer a guest question, or if it encounters an "auto-flag" message type, it will pause Autopilot and flag the conversation for host review.

***

## The "Needs Attention" Flag

When your AI is unable to answer a guest message by itself, it will flag the conversation as **Needs Attention** and automatically filter it into the Needs Attention folder in your Cortado Inbox. 

There are two reasons your AI will flag a message as "Needs Attention"

<details>

<summary><span data-gb-custom-inline data-tag="emoji" data-code="2753">❓</span>AI lacks information to respond</summary>

In order to prevent AI "hallucinations" (made-up responses), your AI will always flag guest queries that exceed the information you have provided it.

For example, if a guest asks "What's the garage code?" your AI will check its [**sources**](/cortado-ai-docs/getting-started/understand-your-ai/what-does-my-ai-know.md) for the garage code. If nothing is found, the AI will route the conversation to you and flag it as "Needs Attention."

*<mark style="color:purple;">When you encounter flagged questions like this one, it is a great time to update your Listing Info Packets to include missing information!</mark>*

</details>

<details>

<summary><span data-gb-custom-inline data-tag="emoji" data-code="2757">❗</span>AI encounters "auto-flag" subject material</summary>

As an added precaution, your AI Agents will *always* flag the following types of requests for human review:

**Auto-flag message types:**

* Emergencies
* Complaints
* Maintenance Requests
* Cancellations or reservation change requests (i.e., change to reservation dates or party size)
* Early check-in or late check-out requests
* Questions about AI
* Special requests requiring human action (e.g., "Can you leave a bottle of champagne in my room?")

</details>

{% hint style="warning" %}
Once a conversation is marked as "Needs Attention" your AI Autopilot will pause itself and will not respond to new guest messages until the conversation has been manually "Resolved."
{% endhint %}

***

## Special Case: Guest acknowledgements & reactions

To cut down on excessive responsiveness and create a more human-like pace of interaction, your AI Agents will not respond to basic guest acknowledgements and reactions.

<details>

<summary><strong>Example Acknowledgements &#x26; Reactions</strong></summary>

The following would NOT receive a response from your AI Autopilot, nor would they be flagged for host intervention:

* Simple acknowledgements
  * "OK"&#x20;
  * "Thanks"&#x20;
  * "No problem"
* Confirmations
  * "Awesome!"
  * "Wonderful!"
  * "Sounds good"
* Positive reactions
  * Guest reacts 😊/❤️/👍/👏/🤣 to your message
  * Guest loves/likes/laughs at/emphasizes your message

</details>

{% hint style="info" %}
When processing a guest acknowledgement or reaction, your AI will not respond and Autopilot will remain on to answer other incoming guest messages.
{% endhint %}