"Do you have anything left for the long weekend?"
It's the question that comes in most often via WhatsApp at the reception of a small hotel. And it's also the one that's answered worst: it arrives on a Saturday at eleven at night, when no one is looking at their phone; it arrives in August mixed in with forty others; or it arrives and gets answered from memory, with a "I think so, let me check" that never gets a second part.
Each of those questions is a booking half a meter from closing. The guest already knows the dates, already knows how many they are, and has already decided they want to go. They just need to know if there's room. If they don't get an answer within a few minutes, they open Booking and ask someone else the question.
Generic AI assistants don't solve this, and it's worth saying clearly: a language modelLLM (Large Language Model)Large language model, the foundation of today's chatbots doesn't know whether you have a double room free on August 14th. It can write an impeccable message saying there is one. And if there isn't, you've just sold something you don't have.
NAiOS Hotel: real inventory, inside the conversation
NAiOS Hotel is the platform's new module that connects the accommodation's booking engine with the NAiOS AI agentsAI AgentsSystems that execute multi-step tasks without constant supervision. From then on, when someone asks via WhatsApp or the website chat whether rooms are available, the agent actually checks before answering.
It's designed for accommodations that already work with a booking engine —the module comes with native support for RoomCloud— and for all sizes: independent hotels, guesthouses, hostels, tourist apartments, rural houses and small chains. A single account can have several establishments connected, each with its own engine, identifier and profile.
The rule that holds everything together: there is, there isn't, or I don't know
Most systems of this kind work with two answers: there's availability or there isn't. We designed the module with three, and that third one is what makes the difference.
If the booking engine is slow to respond, if the date falls outside what the hotel has published, or if something fails along the way, the agent does not say it's fully booked. It says it can't confirm it and refers you to reception.
It seems like a minor detail and it's the opposite. An assistant that answers "sorry, we're full" without having checked is costing the hotel bookings, one by one and silently, without anyone ever noticing. The entire module is built to defend that distinction, and there are automated tests that prevent it from being lost in future versions.
A real case that taught us its importance: a hotel had its calendar loaded only through the end of September. For the booking engine, October and November weren't "closed": they simply didn't exist yet. A naive system would have answered "no availability" for the whole autumn. Ours distinguishes the two situations and answers that it can't yet confirm those dates.
What the agent knows how to do
The module puts five capabilities in the hands of the WhatsApp and chat agent, and uses them only when the conversation calls for it:
- Check availability for specific dates and a number of guests, with the price for the entire stay —not per night— and the minimum stay if there is one.
- Propose rooms that fit that occupancy, sorted by how well they match and by price. If it's five people, it proposes the combinations that fit; it doesn't offer a double.
- View the calendar of available days around the requested dates, so it can suggest alternatives when what's requested isn't available.
- Provide accommodation information: the real capacity of each room type, square meters, beds, bathrooms, services and extras with their price.
- Send the photo of a specific room when the guest asks for it. Over WhatsApp it arrives as an image, not as a link pasted into the text.
- Generate the booking link from the hotel's own engine, with the dates, occupancy, language and direct channel code already set. The guest just has to pay.
That last point deserves a note: the link leads to the hotel's direct channel, not to an intermediary. A conversation that starts on WhatsApp ends on the hotel's own website, with no commission in between.
Reply in the guest's language
The agent replies in the language it's written to and generates the booking link in that same language. For a coastal property, where messages come in Catalan, Spanish, French and English during the same afternoon, this is no mere embellishment.
What you couldn't sell
Every inquiry that ends without availability is logged: which dates were requested, for how many people and how many times. That list is, literally, the demand the establishment couldn't meet. It helps adjust prices, review quotas or decide which weekend it's worth not closing.
Frequently asked questions live separately
Wifi, breakfast, parking, whether pets are allowed or what time check-in is aren't availability: they're hotel knowledge, and they live in the NAiOS Knowledge Base. The module provides the only thing the AI can't know on its own —whether there's room— and leaves the rest where it already worked.
With one important nuance we learned while testing: the data the hotelier hasn't filled in yet is explicitly marked as not recorded, so the agent says "I'll confirm that with reception" instead of making it up. An assistant that invents free wifi generates a complaint at check-in.
Level 1: without touching anything in your RoomCloud account
Here's the part that surprises the hoteliers we've spoken to the most: to get started you don't have to configure absolutely anything in RoomCloud. No permissions to request, no additional module to contract, no support ticket to open, no waiting for someone to enable an access for you.
The module works with the same information the booking engine already publishes to any visitor of the hotel's website. All that's needed is what the hotelier already has on hand:
- The hotel identifier in the booking engine.
- The URL of the search tool on the website.
- The direct channel code, if one is used.
You fill in a form, click "Test connection" and, if the engine responds, the module starts syncing on its own. From there it keeps the calendar for the coming months preloaded and checks live when asked about a date it doesn't have fresh. Setup: minutes, not weeks.
It's the fast track to validate the use case: you connect, test with real conversations and decide. And since nothing has been touched in the engine account, there's also nothing to undo if you prefer to stop.
Level 2: with the RoomCloud API, reading and writing bookings
Level 1 answers the question "is there availability?". Level 2 answers the next one: "book it for me".
With a RoomCloud API key —which the hotel requests from its provider and shares with us— the module goes from querying to operating: reading existing bookings and writing new bookings directly into the engine. What that opens up:
- Closing the booking within the conversation, without forcing the guest to jump to the website.
- Checking a booking that has already been made when the guest writes days later asking about their locator, arrival time or the details of their stay.
- Handling modifications and cancellations through the same channel used to book.
- Confirming with engine data, not with an intermediate copy that can become outdated.
The two levels are not mutually exclusive nor do you have to choose from the start: level 1 goes live today and level 2 is added on top when the establishment wants to take the step, on the same setup and the same agent.
Who it makes sense for
For the small and medium accommodation that already receives enquiries via WhatsApp and doesn't have a twenty-four-hour reception to handle them. Independent hotels, guesthouses, hostels, holiday apartments, rural houses and campsites.
If your establishment receives messages every week asking about specific dates, and some go unanswered until the next day, this module is made exactly for that.
How to get started
NAiOS Hotel is activated like any other module on the platform: you connect the establishment, activate the module in the WhatsApp agent's tools and test it. The platform includes a testing environment where you can see, query by query, what the agent asked the engine, what it answered, where the data came from and how long it took — before putting it in front of a customer.
If you run an accommodation and want to see it working with your own inventory, write to us and we'll set it up with you.
