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

# AI Agent

> Use your AI teammate for guest replies, operating questions, meetings, and approved actions

## What this helps you do

The AI Agent helps your team answer questions, draft replies, find records, and complete approved work. It uses the instructions, documents, connected accounts, and permissions you set in AI Hub.

<CardGroup cols={2}>
  <Card title="Answer guest questions" icon="chat-circle">
    Draft replies using approved property, reservation, and policy details.
  </Card>

  <Card title="Find operating details" icon="magnifying-glass">
    Look up reservations, properties, tasks, contacts, reviews, and other records your team can access.
  </Card>

  <Card title="Take approved actions" icon="check-circle">
    Create tasks, add notes, update records, or start workflows when the agent has permission.
  </Card>

  <Card title="Support meetings" icon="video-camera">
    Join supported meetings, take notes, and create summaries when meeting tools are connected.
  </Card>
</CardGroup>

## Before you start

* Set up the agent in **AI Hub**.
* Add clear instructions for tone, rules, escalation, and what the agent should never do.
* Add the documents and skills the agent can use.
* Connect the channels you want the agent to work in, such as Inbox, email, Slack, WhatsApp, SMS, Telegram, or meetings.
* Review action permissions before letting the agent send messages or change records.

## Write clear agent instructions

Good instructions tell the agent how to behave, what to use, and when to ask for help.

Include:

* Your company voice and greeting style.
* The kinds of questions the agent may answer.
* The situations that need a manager, such as refunds, complaints, safety concerns, legal questions, and discounts.
* Rules for sending messages, creating tasks, and changing records.
* Words or promises the agent should avoid.

<Tip>
  Keep instructions short and direct. If a rule matters, write it as a clear sentence, not a paragraph.
</Tip>

## Connect channels

The agent can work in connected channels when your team enables them:

* **Inbox** for guest, owner, vendor, and team conversations.
* **Email** for connected mailboxes.
* **Slack** for approved workspace channels.
* **WhatsApp, SMS, and Telegram** for supported messaging.
* **Meetings** for supported video calls and recaps.
* **Workflows** for repeat steps that include AI decisions or drafts.

## Control permissions

Permissions decide what the agent can read and what it can do.

Use stricter permissions when the action affects money, guest promises, property records, or outside systems. Use approvals when a human should review the action before it happens.

Common permission choices:

* Read property and reservation details.
* Draft replies without sending.
* Send replies after approval.
* Create tasks or comments.
* Start a workflow.
* Update approved records.

## Look up operating details

The agent can answer questions using records it is allowed to read.

Common lookups include:

* **Reservations** - dates, property, guest count, contact details, notes, and available stay details.
* **Properties** - check-in notes, amenities, access instructions, owner details, and custom fields.
* **Availability** - open dates and matching properties when connected reservation data is available.
* **Tasks** - status, assignee, due date, comments, checklist progress, and related property.
* **Reviews** - ratings, comments, response drafts, and related stays.
* **Contacts** - guest, owner, vendor, and prospect details.

<Tip>
  Ask with the record you care about. For example: "Find Maria Lopez's June reservation and draft a reply about parking."
</Tip>

## Use documents, skills, and memory

AI Hub controls the agent's approved knowledge.

* **Documents** are policies, SOPs, property instructions, and reference material.
* **Skills** are step-by-step instructions the agent can follow.
* **Memory** helps the agent remember approved preferences and repeat procedures.
* **Permissions** decide which agents can use which documents and skills.

Use separate agents when different jobs need different rules. For example, one agent may handle guest replies while another supports owner updates or team operations.

## Read Slack files and shared context

When Slack is connected and permitted, the agent can use supported channel messages and files. It can summarize shared text files, describe supported images, and use that context while answering team questions.

Before relying on Slack context:

* Confirm the Slack account is connected.
* Confirm the agent is allowed to read the channel.
* Keep private or sensitive files out of channels the agent can access.

## Use the AI chat sidebar

The AI chat sidebar is the fastest place to ask the agent for help while you work.

<Steps>
  <Step title="Open the sidebar">
    Open the AI chat from the current page or record.
  </Step>

  <Step title="Ask for the outcome">
    Say what you want, such as "Draft a reply," "Summarize this guest thread," or "Create a maintenance task from this message."
  </Step>

  <Step title="Review the plan">
    For multi-step work, the agent can show a checklist so you can see what it is doing.
  </Step>

  <Step title="Approve or edit">
    Review drafts and any pending actions before they go out.
  </Step>
</Steps>

## Review linked records in chat

When the agent references a record, Trellis can show record pills, inline lists, and detail panels. Use these to check the source before you act.

You may see:

* A reservation, property, task, contact, review, or workflow link.
* An inline list, such as several matching tasks or reservations.
* A detail panel with the related record and a breadcrumb trail when you click deeper.

This is useful when you want to verify an AI answer without leaving your current work.

## Review agent activity

Use **Insights > Pulse** to review recent AI activity. Pulse helps managers see what the agent did, where it happened, and whether anything needs attention.

Activity details can show:

* The conversation, task, property, or reservation involved.
* The action the agent took.
* The result or error.
* Any approval that is still waiting for review.

## Meetings

When meetings are connected, the agent can join supported calls, take notes, and prepare summaries. Use this for owner calls, team handoffs, vendor discussions, and operating meetings where Trellis context is useful.

Before enabling meeting joins:

* Which calendar or meeting account is connected.
* Which meetings the agent may join.
* Who receives recap emails.
* Whether meeting notes should be shared with other agents in your workspace.

Meeting tools can help with:

* Joining supported Google Meet, Zoom, or Microsoft Teams links.
* Taking notes during a call.
* Preparing a recap with action items.
* Linking meeting notes to properties, reservations, or tasks.
* Letting your team ask follow-up questions about the meeting.

Use meeting join rules carefully. Many teams start with external customer or owner calls, then add team meetings only after reviewing how notes are shared.

## Set up each channel

Each channel needs its own setup and permissions.

<AccordionGroup>
  <Accordion title="Email">
    Connect the mailbox, choose what the agent may read or send, and test with a safe draft before allowing sends.
  </Accordion>

  <Accordion title="Slack">
    Connect Slack, choose approved channels, and decide whether the agent may read files shared there.
  </Accordion>

  <Accordion title="WhatsApp, SMS, and Telegram">
    Confirm the phone or messaging account is connected, then review message permissions before allowing guest-facing sends.
  </Accordion>

  <Accordion title="Meetings">
    Connect the meeting account or calendar, set join rules, and decide who receives recaps.
  </Accordion>
</AccordionGroup>

## Use workflows and scheduled work

The agent can help with repeat work when a workflow includes AI steps or when your team asks it to start an approved workflow.

Good uses include:

* Drafting a guest reply that waits for manager approval.
* Creating a task from a conversation.
* Reviewing a booking question before check-in.
* Preparing a daily summary for managers.
* Notifying the team when a sensitive issue needs review.

Keep approvals on workflows that send messages, change records, or affect money.

## How to check it worked

* The agent uses the tone and rules from AI Hub.
* Drafts include the right guest, reservation, property, or task details.
* Any message or record change appears in the related conversation, task, workflow, or activity history.
* Pulse shows the completed action or the reason it needs attention.
* Pending approvals are visible to the right reviewers.

## Common problems

<AccordionGroup>
  <Accordion title="The agent gave a weak answer">
    Add clearer instructions or approved documents in AI Hub. Also check that the agent can read the record it needs.
  </Accordion>

  <Accordion title="The agent cannot send a message">
    Check channel connection, contact details, and the agent's message permissions.
  </Accordion>

  <Accordion title="The agent asked for clarification">
    Answer the question in the same conversation. The agent should continue from there.
  </Accordion>

  <Accordion title="A meeting was not joined">
    Check the connected calendar, meeting link, join rules, and whether the agent is allowed to attend that type of meeting.
  </Accordion>
</AccordionGroup>

## Related articles

<CardGroup cols={2}>
  <Card title="AI Hub" icon="sparkles" href="/platform/ai-hub">
    Set up agents, instructions, documents, skills, and permissions.
  </Card>

  <Card title="Inbox" icon="inbox" href="/platform/inbox">
    Use AI drafts in guest and team conversations.
  </Card>

  <Card title="Workflows" icon="workflow" href="/platform/workflows">
    Add AI steps to repeat processes.
  </Card>

  <Card title="Insights" icon="chart-line" href="/platform/insights">
    Review AI activity in Pulse.
  </Card>
</CardGroup>