Skip to content

Hospitality

Bookings by phone and WhatsApp, with special requests left to your team

We build voice and WhatsApp booking assistants for restaurants, hotels and event venues. They say they are AI, book against your live reservation system or PMS, read every detail back before writing it, and pass allergies, large parties, deposits and complaints to a person with the conversation attached.

Who it is for

Who we build for

Restaurants, hotels and event venues: the same building blocks, pointed at the problem each kind of venue actually has.
  • Independent restaurants and cafés

    The phone rings hardest during service, when the host is seating walk-ins and nobody can take a booking properly.

    We build first · A phone booking agent with a read-back and a WhatsApp confirmation.

  • Multi-outlet restaurant groups

    Each outlet has its own hours, sections and rules, and the central reservations line spends its day looking them up.

    We build first · One agent that routes by outlet, with a rules table per outlet updated in one place.

  • Boutique and independent hotels

    Room enquiries arrive on WhatsApp in several languages at all hours, while the rates sit in a PMS the night team rarely opens.

    We build first · A WhatsApp enquiry assistant that quotes only the rates your PMS returns.

  • Event, banquet and private dining venues

    Large bookings need a person, but the first call goes on the same basics: date, headcount, budget and dietary needs.

    We build first · An enquiry intake that gathers the brief and books a callback with the events manager.

Signature workflow

Follow a booking call

Pick a workflow and step along it: who is speaking, which system is read or written, and where a person takes over. Every line is Sample data.

Phone · An evening call for a table of four: a full first choice, an allergy note and a WhatsApp confirmation.

Sample data

Step 1 of 5: Answer and disclose

  1. The agent says it is the restaurant’s AI booking assistant and offers a person: the host stand during service, a callback after hours.

    Sample data
    Agent: “Good evening, reservations. I’m an AI assistant. I can book a table or put you through to the team.”
    System
    Phone line via Twilio
  2. One question at a time, with names and dates repeated back, because a misheard date is the costliest mistake on the line.

  3. Live availability for that party size, time and section, including turn times. A full first choice gets up to two real alternatives, never a guess.

  4. Occasions and dietary needs go on the booking for the kitchen. The agent never says a dish is safe; the team confirms options.

  5. The booking is read back and written only after the caller says yes, then confirmed on WhatsApp or by text.

Packaged workflows

What we build, packaged

Each package is a fixed-scope build with a named owner on your side. Start with one; they share the same foundation, so the second takes less effort than the first.
  • Phone table booking agent

    Answers your bookings line, says it is AI, books against live availability with a read-back and transfers to the host stand whenever a caller asks.

    • AI disclosure and a transfer on every call
    • Read-back before any booking is written
    • WhatsApp or text confirmation

    Owned by Restaurant manager

    AI voice agents
  • WhatsApp confirmations and reminders

    Template messages to opted-in guests that confirm, remind on the day and let them change or cancel in one reply, releasing the table to the waitlist.

    • Approved templates and opt-in records
    • Confirm, change and cancel replies
    • Waitlist offers

    Owned by Front-of-house lead

    Workflow automation
  • Room enquiry assistant

    Answers room enquiries on WhatsApp and the website with the rates your PMS returns, sends a prefilled booking link and hands anything unusual to the front office.

    • Live rates and availability, read only
    • Languages switched on after review
    • Front-office handoff with the thread

    Owned by Front office manager

    AI chatbots
  • Events and large-party intake

    Collects the brief for groups and private dining, books a callback with the events manager and sends a deposit link only after the manager approves the quote.

    • Party-size rules per outlet
    • Enquiry written to your CRM or events sheet
    • Deposit link after approval

    Owned by Events manager

    AI agents

Operating context

The rules every booking flow is designed around

Booking assistants handle names, phone numbers, dietary needs and sometimes money. We design for these constraints from the first workflow, so your team and advisers can check them before launch.
  • Card payments stay off the call

    The agent never takes card numbers by voice or in chat. Deposits and guarantees go through a payment link from your Stripe account, so card data stays with the payment provider, out of call recordings and out of our systems.

  • AI disclosure and call recording

    Every call starts with the agent saying it is AI and offering a person. Whether calls may be recorded, and what callers must be told first, differs by country and sometimes by state; your counsel confirms the wording for each market.

  • Guest data and privacy law

    Guest details stay in your reservation system or PMS, and conversation logs are kept only as long as your policy sets. We design for India’s DPDP Act, UAE data protection law and GDPR for European guests; your counsel confirms which apply.

  • WhatsApp messaging rules

    Business-initiated messages such as reminders go only to guests who opted in and use templates Meta has approved. Free-form replies are sent inside the customer service window that opens when a guest writes to you.

Shipped by default: AI disclosure on every call, a transfer to a person at any point, read-back before any write, no card data on the line and a log of every booking change. Legal sign-off for each market stays with you and your adviser.

Systems connected

What it connects to

Tools from our build stack, or your own systems through their APIs. The proposal lists every object the workflow reads and writes.
Your reservation or table-management system
Live availability, turn times and sections; bookings written through its API after the read-back.
Your PMS or channel manager
Room availability and public rate plans, read only by default.
Twilio
The phone line for the booking agent, transfers to the host stand and text confirmations.
WhatsApp Business Platform
Enquiries, confirmations and reminders on a number you own, with opted-in guests only.
Stripe
Deposit and guarantee payment links, so card details never pass through the agent.
n8n
Runs confirmations, reminders, waitlist offers and retries, logging every run.
Supabase
The call and booking log, the rules table per outlet and the audit trail of changes.
OpenAI, Anthropic Claude or Google Gemini
The language model, chosen per language after test calls with fluent speakers.

Typical targets

What we plan for, and why

Typical targets are planning ranges we use when scoping, each with its basis. They are not results from a client project; your own baseline decides the real number.
  • Typical target

    40–60%

    Routine booking calls completed without staff

    Basis: Covers new bookings, changes and cancellations; clinical, group, complaint and other complex calls always go to a person. Confirmed against one week of your call log.

  • Typical target

    10–25% fewer

    No-shows on reminded appointments, against your own baseline

    Basis: Moving from no or manual reminders to automated reminders with one-tap rescheduling; reminders reach no-shows caused by forgetting or a clash. Measured against at least four weeks of your own no-show rate before launch.

  • Typical target

    Every enquiry, by design

    Room enquiries answered with a live rate or a handoff

    Basis: The assistant quotes only rates the PMS returns; when that lookup fails, the enquiry goes to the front office instead of getting an estimate.

Concept build

See a worked example

Concept buildMultilingual Guest Concierge for a Boutique HotelGuest messaging after the booking: answers from the hotel guide in tested languages, requests filed as tickets with the room number and complaints sent to the duty manager. A Concept build, not a client project.Open the concept build

Worked examples

Voice concept builds

Concept builds are worked examples, not client work.

See all concept builds
Concept build

AI Voice Agents

After-hours Arabic and English Call Line for a Dental Clinic

A concept build for a voice agent that answers a dental clinic’s phone after hours in Arabic and English and books routine visits into the clinic calendar. It sends an SMS confirmation, moves callers who describe urgent symptoms straight to the on-call dentist, and never gives clinical advice.

  • Healthcare (dental)
  • UAE market
  • Oct 2026
Concept build

AI Voice Agents

Driver Check-in Voice Agent for a Regional Fleet

A concept build for an outbound voice agent that makes routine ETA check-in calls to truck drivers and logs each answer in the fleet’s TMS. It calls only when the truck is parked or at a check-in window the driver chose, and it alerts a dispatcher straight away when a driver reports a breakdown or a delay.

  • Logistics
  • US market
  • Oct 2026

FAQ

Hospitality: questions buyers ask

Does the booking agent tell callers it is an AI?

Yes, in its first sentence, and it offers a person straight away. During service a caller who asks is transferred to the host stand or front desk; after hours the agent takes a callback request, which the team sees on its opening list.

Which reservation systems and PMSs can it book into?

Any system with an API for availability and bookings, confirmed during scoping before you commit. A pilot can also run on a shared calendar or booking sheet. We never keep a second book of our own that your team would have to reconcile with the real one.

How does it handle allergies and dietary requirements?

It records them on the booking for the kitchen and tells the caller the team will go through suitable dishes. It never says a dish is safe or allergen-free, even when the menu lists ingredients, because only the kitchen knows today’s recipes and handling. Callers who want to discuss it now are transferred.

Can it take deposits or card details for large bookings?

Never by voice or in chat. If your policy requires a deposit or guarantee, the workflow sends a payment link from your own Stripe account once the booking or quote is approved, and the booking stays provisional until paid. Refunds stay with your team.

What happens on a busy evening when several people call at once?

The agent can take several calls at the same time, up to the limit set by your telephony plan and the concurrency we configure. Calls above that limit go to a callback request rather than an engaged tone, and the host stand sees every booking the agent writes as it happens.

Can guests book in Arabic, Hindi or other languages as well as English?

Yes, language by language. Each language is switched on only after test calls with fluent speakers pass, including names, dates and spelling read-backs. A caller speaking a language that has not been tested is offered a transfer, so nobody is booked by an agent that half understands them.

We are a hotel, not a restaurant. What does it do for rooms?

It answers room enquiries on WhatsApp and your website with the rates your PMS returns, sends a prefilled booking-engine link and hands unusual requests to the front office. Messaging guests during their stay is a separate workflow, shown in the guest concierge concept build.

Next step

Have a workflow in mind? Let’s scope it together.

Book a free 30-minute strategy call. Bring one process you want to automate and we will talk through scope, integrations and a sensible first step.

Free 30-minute call · no commitment

Send an RFP or security questionnaire