Personal service at scale, from the first message to the next stay.
Every inquiry answered in minutes, in the guest’s language. Every preference remembered. Every offer made personal. Watch the AI read a guest message, look up the hotel’s data, draft a reply in the guest’s language, and wait for the team to approve. Three recorded conversations with a demo resort, replayed step by step.
Start here
The guest who asks the hotel first.
A guest writes on LINE, in Thai, about a sea-view room for three nights. The AI checks rooms and rates, applies Thai politeness and the rule for quoting prices with VAT, drafts a complete reply with the price, and the front desk approves it.
Front desk
Every channel, one inbox, a finished draft. LINE, WhatsApp, Messenger, Instagram and emails from online travel agencies such as Booking.com and Agoda. Staff review a draft instead of researching and typing. Overnight inquiries are drafted and waiting when the shift starts.
Reservations
More direct bookings, less commission. A new guest who writes to the hotel gets a fast, complete answer and a perk chosen for them, so they book here instead of on an agency site. A return guest gets a personal follow-up before the next trip, so they book here again. Every booking made this way is the hotel’s own, with no agency in between.
Marketing
Personalised service for each guest. Direct conversations and stays build a guest record the hotel owns: preferences, past stays, which offers convert. On top of it, AI-powered recommendations: tailored offers, upgrades, add-ons and member perks, matched to each guest and suggested inside the conversation.
Two more conversations
The guest who already booked.
A guest arriving tomorrow writes on WhatsApp, in Simplified Chinese, to arrange an airport pickup. The AI looks up the transfer options, applies the notes for Chinese guests and the ferry timing, drafts a confirmation in Chinese, and the team approves it unchanged.
The request a person finishes.
A company asks on WhatsApp, in English, about a 30-room retreat with meeting space. The AI recognises this is beyond what it may quote, drafts a reply that collects what the sales team needs, and a person edits it before it goes.
What the AI knows
Built by Tres, ready on day one
How Thai, Chinese and international guests expect to be addressed. Polite particles and honorifics, the Buddhist calendar and holidays, simplified or traditional Chinese script, the different tone of LINE, WhatsApp and email. Checked by a Thai language reviewer and kept current by Tres.
Thai, English and Chinese run today. Any further language a hotel wants, such as German, Russian or Japanese, can be added.
Built with the hotel’s team
Rooms, rates and availability from the property system or channel manager. Services, transfers and house rules. Brand, standards and tone of voice, taken from the hotel’s guidelines and learned from every edit the team makes.
In the replays below you can open every lookup the AI made and see the answer it got.
The AI drafts. The team approves.
Nothing reaches a guest on the AI’s own. The team approves, edits or hands over. Every draft, edit and send is logged, and each edit teaches the hotel layer. It keeps learning: staff edits, guest replies and feedback after checkout show what works and what is missing.
In a pilot the hotel can choose routine replies that go out without review. That is the hotel’s decision, made in Discovery.
Running now on a demo resort. Not yet in a hotel.
Running now: the conversation engine and the cultural layer. It reads a LINE message, looks up the hotel’s data, drafts in Thai, English or Chinese and waits for the team to approve. It runs on a demo resort. The three replays on this page are recordings of the engine answering scripted guest messages, two of them written as WhatsApp messages; the replies were not sent to a guest.
Built in a pilot: the hotel layer with a link to the property system, the email lane for online travel agency inquiries, the handover to reservations, the guest record and the feedback loop. How a pilot runs is set out below.
No hotel runs this live yet. The first one would be first. That is why a pilot is priced as a pilot, and why we ask for honest feedback and a case study, named or anonymous, as the hotel prefers.
A four-week Discovery first. Then three phases, each with a gate.
Discovery fixes the phases, their order, the timing and the price. Everything after it is indicative. Each phase starts only when the hotel decides to continue, and the guest record, the direct bookings and the hotel’s own content stay with the hotel. Once the last phase the hotel takes live is complete, one monthly figure covers maintenance and upgrades, running costs included.
- Weeks 1 to 4
Discovery
We map the systems and inboxes, sit with the front desk, reservations and marketing teams, and see how guest inquiries are handled today. Together we set the baseline for direct bookings, decide which routine replies may go out without review, and collect what the hotel layer needs.
Gate. The hotel receives the scope, a build plan split into phases and a fixed price, and decides whether to build.
- Weeks 5 to 12
Phase 1: LINE live
The first four weeks are a trial run: the AI drafts for part of the LINE inquiries, and the team checks and sends each reply. Rooms, rates and availability come live from the property system. Then every LINE inquiry runs through the system.
Gate. Draft quality, review effort and staff acceptance are measured against the targets set in Discovery. The hotel decides on Phase 2.
- Weeks 13 to 20
Phase 2: all channels
The other channels in use: emails from online travel agencies, WhatsApp, Messenger and Instagram. Reservations handover. The routine replies chosen in Discovery go out without review, once the Phase 1 numbers support it.
Gate. The same measures, per channel, and a review of every automated reply type. The hotel decides on Phase 3.
- Weeks 21 to 28
Phase 3: guest record
Guest record live: return guests recognised, personalised offers, upgrades and add-ons inside the conversation, direct-booking perks, a management view. Optional: the feedback loop, with a thank-you message after checkout.
Gate. Direct bookings against the Discovery baseline, and the state of the guest record. The hotel decides on maintenance and upgrades.
What the AI knows →
The cultural layer, the hotel layer, and the team that approves every send.
The cultural notes →
Honorifics, holidays, script rules, channel tone. Plain text, open to read in full.