Loading...
Your Haven AI answers the guests who booked on Airbnb, Vrbo and Booking.com, right in their booking thread, through the PMS you already use. It replies from your listing, your guidebook and what you have taught it, in the guest’s own language — and hands you anything that is yours to decide.
Response time decides bookings and reviews. A traveler comparing three listings books the one that answers first, and a guest standing at the door does not care that it is two in the morning where you are. Most of what they ask is already written down somewhere — in the listing, the guidebook, the house manual — yet someone still has to find it and type it into the right thread. The usual automations do not help: scheduled templates fire at a set time whatever the guest actually asked, and generic AI replies invent amenities, quote the wrong check-in time, or slip a phone number into an Airbnb inquiry and put the listing at risk.
The inquiry that waits until morning is often the booking that went to the listing down the road.
A scheduled message cannot tell the guest whether the crib is in the closet or the hot tub is heated tonight.
Made-up answers, a promised refund nobody approved, contact details in a pre-booking thread: each one a review or a platform warning waiting to happen.
Hospitable, Guesty, Hostfully, Lodgify or Track, plus Hostex, Smoobu, Beds24 and Uplisting on their native connectors. Haven receives each guest message from your PMS as it arrives (on a scheduled check for Track), alongside the reservation it belongs to.
In Communication → Message automations, set each PMS to Draft for approval or Send automatically, and leave out any listing you want to answer yourself. Everything starts off.
The same assistant that runs your Haven sites reads the listing, its guidebook and your AI knowledge. Questions about the area — the weather, drive times, what is open on a Sunday — it looks up live, knowing where your property is. A burst of three messages gets one reply that covers all three.
It is graded against the guest’s question, and screened for contact details before a booking and for door codes or Wi‑Fi before the guest’s arrival window opens. Anything uncertain is held as a draft.
Refunds, changes, complaints and safety questions are handed to you with what the guest needs and a suggested reply, by email and text — and the assistant keeps answering the guest’s other questions while that one waits for you. Only pausing the thread yourself takes it out of the conversation.
Follow one message through
Watch a late-night Airbnb question reach Haven through the PMS, get answered from the host’s own listing and guidebook in the guest’s language, and see what happens when the next message asks for a refund.
A guest who booked on Airbnb, Vrbo or Booking.com writes in their booking thread. Your PMS passes it to Haven, even at 11:47 on a Saturday night.
Maya · 3 guests
The 11 pm “is the hot tub on?” and the 6 am “where do we park?” are answered before you wake, and only the questions that need you are waiting.
One assistant for every listing, each answered from its own details, with per-listing control over where it replies.
Replies in the guest’s language without a translation step or a second template library.
Start in Draft for approval and read every reply before it goes. Move to Send automatically listing by listing, once you trust it.
The reason it answers well is that it is not a separate messaging bot — it is the same assistant, knowledge base and grading loop that runs your Haven sites, pointed at your OTA threads.
Context you add fixing a site-chat gap is the context it uses in the next Airbnb thread. One knowledge base covers your sites, guidebooks and OTA messages.
See AI guest chatDoor codes and Wi‑Fi follow the same reveal window your guidebook enforces — the AI cannot leak what the stay has not unlocked yet.
See the guidebookOTA replies land in the same Chats observability as site chat — answered, low confidence, couldn’t answer — so you see exactly where it needed you.
See analytics & replayScheduled messages send the same words to everyone. Generic AI writes something plausible. Haven answers the guest’s actual question from your own listing, and knows which questions are not its to answer.
Useful for the check-in message on day one. No help at all when the guest asks something the template did not anticipate.
Fast and fluent, but answering from what rentals usually have, with no idea which questions a host would never want an AI to settle.
The assistant already trained on your listing and guidebook, checked before every send, private details gated by booking, decisions handed back to you — and every exchange graded so it keeps improving.
Next step
Connect your PMS, start in Draft for approval, and watch the replies write themselves from your own listing.