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

# Hotels & restaurants

> Reservations, guest messaging, reviews, and supplier coordination — in a service that never stops — with comps, refunds, and contracts always staying human.

It's 8:30pm on a packed Friday. WhatsApp has seven booking requests at once, a
family wants to move its check-in, a guest arriving tomorrow still hasn't heard
anything about their time, a two-star Google review has gone unanswered for
three days, and the produce supplier never confirmed the morning delivery. The
dining room is full, the front desk is on the phone — and all of these landed
together, at the worst possible hour.

**What leaves your plate:** taking and confirming reservations, reminding
whoever's about to arrive, triaging and drafting review replies, and chasing
supplier deliveries all start happening on their own — with context, within
budget, and with **comps, refunds, discounts, and contracts always waiting on
a human**.

## How it works, in one line

Hotels and restaurants are **relationship volume on top of rules that already
exist**: availability, cancellation policy, the menu, the supplier contract.
The agent invents nothing — it reads the rule in the
[Company Brain](/en/features/brain), acts within it, and only decides alone
when the action touches neither money nor a commitment.

```mermaid theme={null}
flowchart LR
    A[Trigger<br/>booking request, guest arriving,<br/>review, schedule] --> B[Agent pulls context<br/>from Company Brain + CRM + availability rules]
    B --> C[Decides + uses tools<br/>WhatsApp, CRM, Routines, Calendar]
    C --> D{Touches money<br/>or a<br/>commitment?}
    D -->|No| E[Confirms / reminds / drafts / chases]
    D -->|Yes| F[Proposes + calls a human<br/>with full context]
    E --> G[Becomes memory<br/>the next guest is already recognized]
    F --> G
```

## Concrete flows

### 1. Booking over WhatsApp — taken and confirmed from availability rules

**Trigger:** a "table for four Saturday at 9pm?" or a room request lands by
[WhatsApp](/en/integrations/whatsapp) or [chat](/en/features/chats).

The agent identifies the guest, cross-checks the request against the
**availability rules** living in the [Company Brain](/en/features/brain) — seats
per slot, minimum notice, group policy, blocked dates — and checks the calendar
through your booking-system integration or
[Google Calendar](/en/integrations/composio). If there's room, it **confirms on
the spot**, logs the reservation as an [activity in the CRM](/en/features/crm),
and returns the confirmation in the guest's channel and language. If there
isn't, it offers the nearest slot that fits the rules. Nothing is guessed: what
it answers comes from what's written.

When the request falls outside the rule — a large group, a special request, a
blocked date — it stops and **hands off to the front desk** with everything
chewed: who they are, what they asked, what the policy says.

<Note>
  The win isn't "a bot that accepts everything." It's that the most common
  booking request — the one that fits the rule — stops pinning the front desk
  to the phone, and a person only steps into the case that truly needs
  judgment.
</Note>

### 2. Pre-arrival messaging — the one that cuts no-shows

**Trigger:** a [routine](/en/features/rotinas) the agent schedules for itself —
it sweeps whoever has a reservation in the next 24 to 48 hours.

On fire, [Marcus](/en/agents/marcus), the communications specialist, pulls from
the [CRM](/en/features/crm) who's arriving, drafts the pre-arrival message —
time, address, directions, what to confirm, the house's tone from the
[Brain](/en/features/brain) — and sends it on the guest's channel
([email or WhatsApp](/en/features/outbound)). It asks for a one-tap
confirmation; whoever confirms keeps the reservation; whoever says they're not
coming **releases the slot** back to the calendar. Replies flow back into the
CRM as activity.

<Warning>
  Pre-arrival messaging is external communication at volume. Keep **human
  approval before each send** at first — the agent prepares the whole queue,
  you review the tone and release it. Autonomy grows like a ratchet: once it's
  reliable, you let the batch go. Any **comp, upgrade, or discount** that goes
  into the message stays a proposal for a human to approve.
</Warning>

### 3. Review triage — a draft ready to approve

**Trigger:** a review pings on Google, on TripAdvisor, or on the delivery app.

The agent reads the review, **classifies** it — praise, a food complaint, a
service complaint, a booking problem — cross-checks it with the guest's history
in the [CRM](/en/features/crm) — the same
[support-and-relationship](/en/use-cases/people) pattern that holds a returning
guest — and **drafts the reply** in the house's tone:
thanks the praise, acknowledges the problem, points to the next step. The draft
sits **waiting for your approval** before publishing. A serious review opens a
[task](/en/features/boards) assigned to the right owner — kitchen for the food,
floor for the service — with the whole case on the card.

```mermaid theme={null}
flowchart TD
    A[New review] --> B[Agent reads + classifies<br/>+ cross-checks history]
    B --> C{Type}
    C -->|Praise| D[Drafts a thank-you → human approves]
    C -->|Operational complaint| E[Task → right owner + draft]
    C -->|Recovery that costs money| F[Proposes comp/discount → human approves]
    D --> G[Becomes memory]
    E --> G
    F --> G
```

The agent runs everything up to the money line. **Comps, vouchers, refunds, and
any recovery that commits money it proposes — a person releases.**

### 4. Supplier coordination — order placed, delivery chased

**Trigger:** a restock [routine](/en/features/rotinas), or a stock level that
drops below the minimum recorded in the [Brain](/en/features/brain).

The agent builds the restock list from consumption and the minimums the Brain
knows, prepares the order for each supplier, and — **with your approval** —
sends it on each one's channel ([email or WhatsApp](/en/features/outbound)).
Then it **follows up**: it cross-checks promised against received, and if the
morning delivery didn't confirm, it chases the supplier and opens a
[task](/en/features/boards) so the front desk knows before anything runs short.
[Scout](/en/agents/scout) researches what's public — an alternate supplier, a
delivery window — when you need a quick option.

<Info>
  **Why this matters:** delivery confirmation is the job nobody remembers to do
  until the ingredient runs out mid-service. The agent chases the supplier as a
  routine and flags the gap **before** it happens — not after.
</Info>

## What stays human

<Warning>
  The agent **proposes**, never decides alone, whenever the subject is:

  * **Comps, refunds, and any money** — courtesy, voucher, chargeback, bill
    adjustment. Every financial move goes through approval.
  * **Discounts, upgrades, and comps** — anything that changes the final value
    of the stay or the check.
  * **Service recovery that commits money** — the free night, the dinner on the
    house, the apology credit.
  * **Supplier contracts** — closing, renewing, or changing terms with whoever
    stocks the house.
  * **A commercial proposal to a customer** and heavy external, legal, or
    regulatory communication.
  * **Governance decisions and destructive actions** — deleting records,
    canceling reservations in bulk.

  In those cases the agent prepares everything — the draft, the context, the
  recommendation — and **waits for a person to approve.** Every action lands in
  the auditable trail, with author and justification, and spend is capped by
  [Stars](/en/billing/stars) before the call ever happens.
</Warning>

## Where to start

<Steps>
  <Step title="Connect the channel and the calendar">
    Wire up [WhatsApp](/en/integrations/whatsapp) and connect your booking
    system or [Google Calendar](/en/integrations/composio). That's where the
    agent reads availability and confirms the reservation without depending on
    the front desk.
  </Step>

  <Step title="Teach the rules to the Brain">
    Upload your availability rules, cancellation policy, the house's tone, and
    your per-supplier stock minimums into the
    [Company Brain](/en/features/brain). Without it, the agent has nothing to
    pull a first confirmation from.
  </Step>

  <Step title="Schedule pre-arrival">
    Create a [routine](/en/features/rotinas) that sweeps reservations in the
    next 24 to 48 hours and let the agent assemble the pre-arrival queue — with
    human approval before each send at first.
  </Step>
</Steps>

## Next steps

<CardGroup cols={2}>
  <Card title="Operations" icon="gears" href="/en/use-cases/operations">
    The high-volume support and triage pattern that holds up a service that
    never stops.
  </Card>

  <Card title="WhatsApp" icon="whatsapp" href="/en/integrations/whatsapp">
    The channel where the booking and the guest happen — request, confirmation,
    and pre-arrival in the same place they already are.
  </Card>

  <Card title="Routines" icon="clock-rotate-left" href="/en/features/rotinas">
    The cron behind pre-arrival and restocking — the agent fires on the right
    day, with nobody remembering.
  </Card>

  <Card title="Native CRM" icon="address-book" href="/en/features/crm">
    Where every guest, reservation, and interaction stays alive — the history
    that makes a returning customer feel recognized.
  </Card>
</CardGroup>