For independent hotel owners, homestay hosts, boutique chains and the manager who runs the WhatsApp

AI agents for hotels and homestays: every guest message answered, every booking confirmed, in the guest’s language.

An agent that answers enquiries from WhatsApp, the website, Google and the OTAs within a minute, quotes from your live rates and availability, holds and confirms direct bookings, sends the pre-arrival message and the directions, handles the in-stay requests that are really tasks for housekeeping or the kitchen, and asks for the review after checkout. Staff keep every judgement about a guest; the agent keeps the inbox at zero.

Thirty minutes, free. You leave knowing which workflow to start with, what a pilot covers and what it costs.

PilotAI Agent Pilot: one workflow, 2–3 weeks, $4,500 (₹3,50,000) fixed. What the pilot includes.

Who this is for: Independent hotels, homestays, boutique chains and serviced apartments with 5 to 80 rooms, enquiries arriving on WhatsApp, Instagram, the website, Google Business Profile and OTA chat, a property management system or a channel manager (or a very good spreadsheet), and a front desk that is also the reservations team, the concierge and the night auditor.

The week this fixes

If this is your Monday, keep reading

The 11 pm enquiry books on an OTA by midnight

A guest asks on WhatsApp whether a family room is free next weekend. The desk is checking in a late arrival. By the time someone replies in the morning the guest has booked the same room on an OTA, and the property pays 15–25% commission for a guest who came to it directly.

The same twenty questions, in four languages, all day

Check-in time, early check-in, breakfast hours, airport pickup, is there parking, do you allow pets, what is the cancellation policy, can you send directions. The front desk answers each one from memory, with small differences, while the guest in front of them waits.

Reviews ask for themselves, which is to say they do not

Happy guests leave without being asked; unhappy ones find the review form unprompted. The property’s rating is written by the second group. Nobody has time to message every departed guest at the right moment.

One workflow, before and after

From a late-night enquiry to a confirmed direct booking and a reviewed stay

Trigger: A WhatsApp message at 10:48 pm: "Hi, do you have a room for 2 adults 1 child from the 14th to the 16th? What is the rate with breakfast?"

Today

  1. GuestMessage sits on the reception phone. Opens an OTA to compare.
  2. DeskReplies at 8:40 am: "Yes available, rate 4,500 plus taxes." No photos, no breakfast answer, no link to book.
  3. GuestAlready booked on the OTA at 11:20 pm, same room, with the property paying commission.
  4. DeskOn arrival: asks for ID, explains breakfast hours, draws directions to the parking. Guest asks for extra pillows at 9 pm; the request reaches housekeeping at 9:40.
  5. PropertyGuest leaves happy. Nobody asks for a review. A month later a guest with a plumbing complaint does leave one.

With the agent

  1. AgentReplies within a minute in the guest’s language: confirms dates and occupancy, reads live availability and the direct rate for a family room with breakfast from the PMS or channel manager, sends two photos and the cancellation policy.
  2. AgentOffers to hold the room for an hour and sends a payment link for the deposit from the property’s gateway. Guest pays at 10:56 pm. Booking written to the PMS; confirmation sent.
  3. AgentTwo days before arrival: directions, parking note, check-in time, ID requirement, and "anything we should prepare?" Guest mentions a toddler; a cot is added to the housekeeping sheet.
  4. AgentDuring the stay: "extra pillows please" at 9 pm becomes a task on the housekeeping board at 9:00, acknowledged to the guest, closed by housekeeping at 9:12. A complaint about hot water goes to the manager’s phone immediately, not into a queue.
  5. AgentDay after checkout: a thank-you in the guest’s language and one link to review. A guest who reported a problem gets the manager’s message instead, and no review request.

What stays with a person

  • Rate decisions beyond the published direct rate: discounts, long-stay deals, group quotes. The agent quotes what the system shows and says "the manager will confirm" for anything else.
  • Any complaint, refund or safety matter: a person, immediately, with the thread attached.
  • The welcome. The agent gets the paperwork and the questions out of the way so that the person at the desk can look up and say hello.
What it saves

Direct bookings, front-desk hours and reviews: a worked example

MeasureTodayWith the agent
First reply to an enquiryNext morning for a third of themUnder a minute, any hour, any language
Front-desk hours on enquiries, per month~25 hours~6 hours
Front-desk hours on routine in-stay requests, per month~20 hours~4 hours
Departed guests asked for a reviewWhoever the manager remembersEvery guest without an open complaint, the day after checkout
Night enquiries that can book directlyNone until morningAll of them, with a deposit link

Worked example. The assumptions:

  • A 24-room property receiving about 300 enquiries a month across WhatsApp, Instagram, the website and Google, a third of them outside desk hours
  • About 5 minutes of staff time per enquiry answered by hand, including looking up availability and typing the rate
  • About 400 routine guest questions and requests a month during stays, about 3 minutes each
  • The agent completes about 75% of enquiries and 80% of routine requests within the rules; the rest reach a person with context

Hours are the small saving. The number that matters is how many of the hundred night-time enquiries book direct instead of on an OTA, at 15–25% commission each, and how the review count and rating move. A pilot measures direct-booking share and review volume against the previous quarter rather than quoting an uplift here.

Put your own numbers in the automation savings calculator, or get a price range from the AI agent cost estimator.

The agents

Usually one to start, three when it is working

Each agent has a small set of tools it is allowed to call and nothing else. That list is the guardrail, and it is written down here so you can see it.

Reservations agent

On WhatsApp, Instagram, the website and Google Business messages, in the guest’s language. Reads live availability and the direct rate, answers the fixed-knowledge questions from the property’s approved facts, sends photos and policies, holds a room, takes the deposit through the property’s payment gateway, writes the booking to the PMS and sends the confirmation. Hands over for groups, special rates and anything it is not sure about.

  • check_availability
  • quote_direct_rate
  • answer_from_property_facts
  • hold_room
  • send_payment_link
  • create_booking
  • handover_to_desk

Pre-arrival and in-stay agent

Sends the pre-arrival message with directions, timings and the ID note, collects preferences and special needs onto the housekeeping sheet, turns in-stay requests into tasks for housekeeping, the kitchen or maintenance with an acknowledgement to the guest, and escalates complaints to the manager the moment they appear.

  • send_pre_arrival
  • collect_preferences
  • create_task
  • notify_department
  • acknowledge_guest
  • escalate_complaint

Review and return agent

The day after checkout, thanks the guest and sends one review link, skipping anyone with an open complaint. Drafts the reply to each new public review for the manager to approve. Reminds past guests about the season they stayed in, once a year, to those who opted in.

  • list_departures
  • check_open_issues
  • send_review_request
  • draft_review_reply
  • send_return_offer

Guardrails

  • Rates come from the PMS or channel manager, never invented, and the agent never undercuts the published direct rate or breaches rate parity with the OTAs; anything off-rate goes to the manager.
  • Property facts come from one approved list the owner maintains (timings, policies, amenities, distances); the agent says "let me check" for anything not on it rather than guessing.
  • Complaints, refunds, safety and anything medical go to a person immediately and the agent stops offering reviews to that guest.
  • Messaging only to guests with consent on record, within WhatsApp’s template and opt-in rules; the return offer goes only to guests who opted in, once a year.
  • Guest IDs and payment details never pass through the agent: the deposit is a link to the property’s gateway, and ID is collected at the desk or through the PMS’s own secure form.
  • Every message and action is logged against the booking, so a dispute has a transcript and the owner can audit a week in ten minutes.

Systems it talks to

  • WhatsApp Business API, Instagram DMs, the website chat and Google Business Profile messages
  • PMS and channel manager: eZee, Hotelogix, Cloudbeds, STAAH, AxisRooms, Little Hotelier, Mews, or the booking sheet you have outgrown
  • Payment gateway: Razorpay, Cashfree or Stripe for deposit links
  • A shared housekeeping and maintenance board (a sheet, Trello or the PMS’s task module); Google Business Profile and OTA extranets for reviews
Proof, and its limits

What we have built, and what we have not measured yet

The clinic booking agent on the demos page is this design with rooms swapped for appointment slots: it reads availability, offers times, confirms and reminds. The reservations agent for a property is a change of nouns and a payment link. We have not yet published a hotel result and will not quote one until a property agrees to publish theirs; the reference builds show the mechanics, and the pilot measures direct-booking share and reviews on your own property.

FAQ

Questions hotels and homestays ask us

Will the agent quote a wrong rate or an unavailable room?

It reads availability and the direct rate from your PMS or channel manager at the moment of the question, and quotes nothing it did not read. If the connection is down it says so and hands over rather than guessing. Rate parity is a rule in its rulebook: it never goes below the published direct rate, and anything off-rate is a manager’s decision.

Which PMS and channel managers does it work with?

Cloudbeds, Mews, Little Hotelier, eZee, Hotelogix, STAAH and AxisRooms have APIs or connectors of varying depth; a property on a booking sheet can start with the sheet and connect the PMS once the pilot has proved it. OTA chat (Booking.com, Agoda, MakeMyTrip) is reachable through the extranet messaging where the OTA allows it; the agent’s strongest channel is the direct one.

Does it work in Hindi, Marathi, Tamil, or for foreign guests?

It replies in the language the guest writes in, including Indian languages and the common European and East Asian ones, and keeps the property’s facts the same in every language. Owners usually review the first week of transcripts in the languages they read and spot-check the rest with staff.

What happens to a complaint?

It goes to a person immediately, by WhatsApp and if configured by call, with the whole thread. The agent acknowledges the guest, tells them who is coming, and stops any automated messaging to that guest, including the review request, until the manager closes the issue.

What does it cost?

The AI Agent Pilot covers the reservations agent on WhatsApp and the website for one property for a fixed fee in two to three weeks, measured on direct-booking share and first-reply time against your last quarter. Pre-arrival, in-stay and review agents are scoped after. The cost estimator gives a range for your setup.

Next step

Tell us about your property

How many rooms, which PMS or channel manager, where the enquiries arrive, which languages. We reply with what a pilot on your reservations would cover.

Or start with a number: the AI agent readiness quiz takes three minutes and gives you a score and a 30-day plan.

Or send the details

Name and email are all we need. The rest helps us reply with something specific.