AI guest messaging for vacation rentals
TravelPal answers your guests' messages around the clock, in the language the guest writes in. Where it is confident, it replies and acts on its own; everything else goes to your team instead of the guest, inside the same conversation.
An illustrative conversation; the guest and unit are invented.
A guest who messages at 3 AM gets an answer at 3 AM.
Replies go out at any hour, because guests do not save their questions for business hours. Messages the AI starts itself, reminders and offers, are held overnight and sent in the morning, on the clock at the property.
There is no language list and no language setting; the reply mirrors the guest's own message, Spanish included, with more detail in guest messages in Spanish and other languages.
Two gates every reply clears before a guest sees it
Every reply clears two independent gates before it reaches a guest, decided message by message. Anything that does not clear both lands in front of your team instead, inside the same thread, where only your team can see it.
Clears both gates
Same thread, visible only to your team.
Subjects you name go to a person
You list topics the AI must never handle alone, in your own words. A reply touching one goes to your team no matter how confident the AI was.
Uncertain replies go to you
The AI scores its own reply against a confidence bar you set from your dashboard. A reply below the bar goes to your team inside the same conversation, and a reply that could not be scored at all is never sent.
It never talks over your team
If someone on your team has already replied in the thread, the AI's pending reply is withheld. Step into a chat from the dashboard and it stands down there too.
Complaints and refund requests stop the AI
A complaint or a request for money back stops the AI in that conversation entirely, and it stays stopped until a person takes over. The full list of guardrails is on when the AI stops.
Underneath both gates sits one rule: grounding. For any question about the unit, the building, fees, or policies, the answer comes from a fact on file, your house manual, your listing content, or a learned answer, or the AI asks your team first.
Anything it cannot answer goes to your team, not the guest.
A reply that does not clear the bar is not thrown away. It is delivered to your team inside the very thread the guest wrote in, so picking it up costs one glance, and your team can send it, edit it, or take the conversation over.
Held, in the guest's own words. When the answer simply is not on file, the AI holds the guest's question for your team exactly as it was written.
The guest is told. The guest hears that the team has been asked, in the same conversation.
Your answer is relayed back. It lands in the very chat the guest wrote in, so the thread reads as one conversation.
The full flow is on what happens when the AI does not know.
You named building work as a subject a person handles, so this reply waits here, visible only to your team, instead of going to the guest.
An illustrative conversation; the guest and unit are invented.
Every reply lands in the inbox you already use.
Messages TravelPal sends go out through your own property management system, never around it. Your team's own inbox shows exactly what each guest was told.
It sells while it answers, too: the mechanics are on upsells and the integration page for the system you run. For the whole loop from a team's point of view, see how TravelPal works for property management companies.
The same reply, as your team sees it.
Guest messaging FAQ
Does it really answer at 3 AM?
Yes. Replies go out around the clock, and a guest who messages at 3 AM gets an answer at 3 AM. Messages the AI starts itself, like reminders and offers, are different: they hold overnight and go out in the morning, on the clock at the property, so guests are answered all night and never woken by a message.
Which languages does it speak?
It replies in the language the guest writes in, Spanish included. There is no language list and no setting to configure. When it relays fixed text such as payment instructions, the sentence is translated and the payment link is left untouched, character for character.
What happens when it does not know an answer?
It does not guess. The question is registered with your team in the guest's own words, the guest is told the team has been asked, and your answer is relayed back into the same conversation. You can save the answer so that question never has to be escalated again.
Can I control what it handles on its own?
Yes. You name subjects that must go to a person, and a reply touching one goes to your team no matter how confident the AI was. You also set the confidence bar from your dashboard; a reply below it lands in front of your team instead of the guest. A complaint or a refund request stops the AI in that conversation entirely, and your team is alerted the moment it happens.
Updated