A fractional AI Support Department for hospitality, 24/7 in 10+ languages.

Pull the last three months of guest comms from a typical boutique hotel group with three to eight properties. The shape is consistent and brutal. Forty percent of the inbound queue is reservation modifications: room changes, date changes, room-type upgrades, extra nights, late check-out requests, early check-in requests. Twenty percent is pre-arrival special requests: airport transfers, dietary requirements, accessibility needs, room preferences, anniversary surprises, baby cot requests. Fifteen percent is on-property questions during the stay: housekeeping requests, restaurant reservations, concierge questions about the neighborhood, billing questions, AC issues, towel requests. Ten percent is restaurant inquiries: menu questions, dietary restrictions, group bookings, private dining requests, time slot availability. The remaining fifteen percent is everything else, including the truly novel questions that need a human.

The other thing the export shows is when the queue lands and what language it lands in. The inbound peak is 9pm to 1am local time, exactly when the front desk is covering the property with one person who is also checking in late arrivals. The language mix for a boutique group in Hong Kong, Singapore, Bangkok, Tokyo, or any cross-border destination runs through fourteen languages in a typical month. Mandarin, Cantonese, Japanese, Korean from the APAC source markets. English from the US, UK, and Australia. French, German, Italian, Spanish, Portuguese from the EU. Thai, Vietnamese, Bahasa from regional inbound. Arabic from the Middle East source markets. The front desk staff are fluent in two, maybe three of those languages on a good shift. The rest get Google Translate replies that read as polite but uncertain, which trains the guest to switch to a brand that handles their language natively.

The math says a well-built support layer should be closing eighty to ninety percent of inbound on its own across email, WhatsApp, OTA messaging (Booking.com inbox, Agoda chat, Expedia messaging), in-app chat, and SMS. The reality is most boutique groups close less than thirty percent of inbound autonomously because the front desk does not have the bandwidth and the marketing person who could run a chatbot does not have the time. The Italian restaurant down the road has no Korean menu support and a TripAdvisor page with three answered reviews out of forty. The boutique group with strong on-property service is leaking guests to chain brands that pay engineers to staff what humans cannot. See AI for Hospitality for the integrated picture across all four functions.

Hospitality is the most multilingual function in the small-business economy and the worst-equipped to handle it. A boutique hotel in Hong Kong takes guests from mainland China, Japan, Korea, Singapore, Malaysia, Australia, the UK, Germany, France, and the US in any given month. A boutique hotel in Tokyo takes guests from Korea, Greater China, the US, the UK, France, Italy, and Australia. A boutique hotel in Bangkok takes guests from China, India, Korea, the UK, Germany, Russia, Australia, and the US. In every case the guest base spans ten-plus languages and the staff covers two or three of those well. The gap is structural and headcount cannot close it.

The agents we deploy speak the languages your guests speak natively. The inbound language is detected on the first word. The reply lands in that language in the brand tone you have spent years setting, with the right formality register for the culture. A Korean guest writing in formal Korean gets formal Korean back. A casual WhatsApp message in Cantonese gets a friendly Cantonese reply that knows what dim sum the hotel restaurant serves. A French guest asking about the wine list at the in-house bistro gets a French reply with the actual wine list information pulled from the F&B knowledge base. Your front desk staff stay focused on the guests in front of them in the languages they actually speak. The queue across every other language closes itself.

The agent is multi-property aware from day one. The guest in Tokyo asking about the rooftop bar at the Hong Kong property gets the Hong Kong property answer. The guest in Seoul asking about the spa at the Singapore property gets the Singapore property answer. The agent knows which property the inquiry maps to, pulls the right operating hours, the right F&B offering, the right amenity grid, the right local context, and answers in the guest language. Reservation modifications get applied to the right PMS instance with the right property code. Special requests get routed to the right on-property GM or concierge with the full guest context attached.

11:47pm, Hong Kong property: a Japanese guest in Tokyo writes in through WhatsApp asking whether the rooftop bar serves dinner. The agent reads the message in Japanese, pulls the rooftop bar operating hours from the F&B KB, confirms dinner service runs until 10pm Sunday to Thursday and 11pm Friday to Saturday, suggests a reservation slot for the guest arrival date, and offers to hold a corner table with city view. The reply lands in formal Japanese in the brand voice in thirty-two seconds. The guest thanks the agent, confirms the table for 8pm on arrival day, and the reservation lands in the in-house F&B system with the guest preferences noted. The cart is saved, the table is booked, and the guest arrives to a corner table with a "welcome back" greeting in Japanese.

3am, Bangkok property: a Korean guest in Seoul wants to extend her stay by two nights and asks in Korean. The agent reads the message in Korean, pulls the reservation from the PMS, checks availability for the extension nights, scopes the rate (the original rate applies for the extension under the group rate code), and confirms the change in formal Korean in the brand voice. The PMS reservation is updated. The guest gets a confirmation in Korean inside the hour. The next morning the front desk has the updated rooming list ready when they walk in. No back-and-forth, no Google Translate uncertainty, no guest defecting to a different property for the extension.

6am, Singapore property: the in-house Italian restaurant inbox has overnight messages in Mandarin, English, French, and Japanese asking about reservations for Friday night, dietary accommodations for a vegan guest, a private dining room booking for a corporate dinner of twelve, and a request to add a candle and rose petals for an anniversary surprise. Each message gets a reply in the source language. The Friday night reservation lands in OpenTable with the table preference noted. The vegan accommodation is confirmed against the menu. The private dining room booking gets escalated to the F&B director with the corporate context attached. The anniversary surprise gets routed to the FOH manager and the housekeeping team with the surprise plan attached.

11am, every property: the GM opens the night queue summary. Eighty-seven tickets came in between 9pm and 9am across the four properties. Seventy-six were closed end to end (reservation modifications, special requests, F&B inquiries, concierge questions). Nine were queued for human review with full PMS context and a draft response attached. Two were flagged as VIP escalations from high-AOV repeat guests. The GMs spend their morning on the nine that needed them, not on triaging the eighty-seven that did not. For the integrated view across all four hospitality functions, see AI for Hospitality. For the cross-industry support context, see AI Support Department and 24/7 AI Customer Support.