How to Automate Hotel Guest Messages, Step by Step

Why this problem costs money

Most of the cost is front desk attention, not message fees. Each booking confirmation or "what time is check-in?" message that a person types by hand takes a few minutes, and it tends to arrive when the desk is busiest. As an illustration only, if a property gets 40 routine questions on a heavy day and each takes three minutes across phone, email and chat, that is two hours of desk time. Replace those numbers with a count from one normal week in your own inbox before you make any decision.

Speed matters to guests. A 2025 analysis by Hotel Tech Report, published in April 2025, found that 82% of text messages are read within five minutes. Cloudbeds, a hotel management software vendor, reports that 73% of travellers prefer messaging over phone calls when they contact hotels, and that they expect replies within an hour. Because that second figure comes from a software vendor, treat it as a vendor view rather than an independent measurement.

Guests also accept automation for simple questions. The 2025 State of Hotel Guest Tech Report, based on a survey of 402 recent hotel guests, found that 70% consider chatbots helpful for simple inquiries. A separate trade report from Hotel Technology News in July 2025 claims that leading messaging platforms can handle 70 to 90% of routine inquiries without staff. That range is a vendor-market claim. HelloShift's own platform data for 2025 is more modest: roughly 60% of the messages hotels sent through that platform were automated, mostly trigger messages such as confirmations and pre-arrival instructions.

The message fees are small. On Meta's rate card as summarised by Kanal in October 2026, a utility message to a recipient in the United States costs about $0.0034, and in the United Kingdom about $0.0220. Blueticks, another vendor blog, lists the US utility rate at $0.004 as of June 2026, so the sources differ slightly. Check Meta's own pricing page before you budget. As arithmetic, 3,000 utility messages a month would cost roughly $10 at the US rate and roughly $66 at the UK rate, before any messaging platform fee. Staff time is almost always the larger number.

Manual work versus automated work

The table below compares how most properties handle these messages today with a setup where the system sends the routine messages and a person checks the exceptions.

The message sequence from booking to checkout

Map the stay as a short chain of messages before you configure anything. A typical chain for a single guest runs like this: the booking confirmation when the reservation is made, the check-in details about a day before arrival, a welcome message with the room number and Wi-Fi details once the room is ready, a mid-stay check-in on the second day, and a checkout note with the checkout time and invoice link, followed by a review request. Sendsquared, a hospitality messaging vendor, describes a similar pre-arrival and in-stay sequence in its 2026 WhatsApp guide for hotels.

The welcome message is the one most worth automating, because most of the "where do I find" questions arrive in the first hour. Front desk guides such as Prostay's operations notes recommend that the system sends the room and Wi-Fi details right after check-in is completed in the property management system, which removes many of those questions.

Which questions an AI reply can answer

An AI assistant does well with questions whose answer is written down and does not change from guest to guest. Good candidates include:

• Check-in and checkout times, and the front desk phone extension or number
• Wi-Fi network name and password, parking rules, and airport or station directions
• Breakfast hours, pool or gym opening times, and house rules
• How to request a cot, extra towels or an early or late arrival
• Payment methods accepted and where to find an invoice

Late checkout deserves a split answer. The AI can explain the policy and take the request, but whether it can be granted depends on whether the room has an arrival that day and whether housekeeping can prepare it. A standard front office test notes that late checkout is possible only when there is no arrival for that room and housekeeping has time, so the decision stays with the desk.

Step-by-step: how to set this up

1. Count the questions. Export two weeks of messages from email, WhatsApp, the booking channel and the phone log. Group them into the topics above and rank them by frequency. Keep the top twenty for the first version.
2. Choose the channel. The free WhatsApp Business app is fine for one phone and a few messages a day, but it has no automation or shared inbox at scale. For automation, you need the WhatsApp Business Platform, usually through a Business Solution Provider, which charges a platform fee on top of Meta's per-message charges. Ask each provider for its current pricing; do not assume a figure from a blog.
3. Register the message templates. Every message you start outside the customer service window must use a template approved by Meta. Templates fall into marketing, utility or authentication categories, and the approval review can take up to 24 hours. Booking confirmations and check-in details are utility templates; keep promotions in a separate marketing template.
4. Connect the booking data. Many property management systems, including Cloudbeds, Mews and Oracle Opera, have messaging or API options; check with your vendor which ones your version supports. Connect the booking date, room number, arrival time and language fields so the templates fill in automatically.
5. Write the policy sheet. Put every routine answer in one spreadsheet, such as Google Sheets or Excel: question, answer, and the owner who approves changes. The AI assistant should answer only from this sheet. When check-in times or Wi-Fi details change, you update one row, not five documents.
6. Set the escalation rules. Decide in writing which topics go straight to a person: refunds, complaints, disability or medical needs, VIP or group arrivals, and anything about ID checks. Meta's business policy requires that automated replies inside the 24-hour window come with clear and direct paths to a human.
7. Handle ID checks separately. Some properties use optical character recognition (OCR) to read identity documents and fill in registration fields. Whether that is allowed, what must be kept and for how long depends on the registration and data protection rules where the property operates. Confirm this with your local authority or a lawyer before you turn on any document capture, and keep the final check with a trained person.
8. Connect the accounting record. Charges for extras, late checkout fees and invoices should flow into the accounting software you already use, such as Xero or QuickBooks, so the message system does not create a second set of numbers.
9. Run a two-week trial. Let the system draft replies while staff approve each one, then switch to automatic replies for the routine topics only. Read the escalations every morning for the first month.

What breaks, and how to prevent it

The most common failure is the 24-hour rule. Once a guest has been silent for 24 hours, the business can send only a pre-approved template until the guest writes again. If the system sends a free-form reply after that point, the message will fail. Build a template for every message that may go out later than a day after the last guest contact, such as a late schedule change.

The second risk is stale data. If the booking system and the messaging system disagree about the arrival time, the guest gets the wrong instruction. Run a daily check on new and changed bookings for the first month, and compare a sample of sent messages against the reservation.

The third risk is double sending. A guest booked through an online travel agent and then contacted directly can receive two confirmations. Pick one source of truth for each booking channel and turn off the duplicate trigger.

The fourth risk is template quality. Meta reviews templates continuously, and a high rate of guests blocking or reporting a business can lower its sending limit. Send only the messages a guest asked for or booked, and make the opt-out wording clear in each template.

Finally, language mismatch can cause confusion. If the AI replies in a different language from the guest, the answer may be correct but hard to act on. Set the default language on the reservation and let the guest choose another in their first message.

When not to automate, and what the front desk still handles

Automation is the wrong tool for some guest contact, and the desk should keep these cases. VIP arrivals, guests with special access needs, families with a medical concern, guests who ask for a manager, and any complaint about cleanliness, noise or safety need a person. So do ID-check exceptions, such as a guest without a standard document, a name mismatch, or a visitor whose registration details do not match. Groups with custom terms, corporate rates and refund requests are also better handled by hand.

Automation also does not pay off in every property. A small guesthouse where the owner answers every message personally, in a few minutes, may gain little from a platform and a policy sheet. The case for automation is strongest where the same five questions arrive every day, where messages come in overnight, or where more than one person answers the same inbox.

If the change affects someone's hours or duties, agree it with that person first. Notice, consultation and redundancy rules apply in many countries, so check the position with a lawyer or your labour authority before you change roles.

This explains the general approach in plain terms; it is not legal or tax advice for your case.

What it typically takes: time, not price

The first setup is mostly preparation. Expect the question review and the policy sheet to take the most effort, because the answers must be correct before the system sends them. Template approval adds a short wait, and connecting the booking system depends on what your property management system allows. For planning, most owners should set aside a few weeks of part-time work for a first version covering one property, then a further two weeks of reading escalations and correcting answers. This is a planning assumption based on the steps above, not a measured figure.

After launch, the ongoing work is smaller but steady. Someone should update the policy sheet when rates, times or house rules change, check escalations daily, and review the template list every few months. Whether the setup saves money or time depends on your own message counts and staffing, which is why measuring one normal week before you start matters more than any published average.