What happens when the AI does not know
TravelPal's AI does not guess, and it does not stall the guest. The question goes to your team in the guest's own words, the guest is told you have been asked, and your answer comes back into the same conversation, remembered for next time if you choose.
An illustrative conversation; the guest and property are invented.
How escalation works, in four steps
The question is held
When the knowledge lookup finds nothing, even after a retry with different wording, the AI registers the question exactly as the guest asked it. Only then does it tell the guest the team has been asked, so the sentence is true when the guest reads it.
Your team hears at once
The moment a question is held, your team is told two ways at the same time: a message in your team channel and an email to your inbox. The alert carries the guest's name, the property, and the question itself.
You answer once
The question waits as a card in your dashboard inbox. One person answers it once; a double submit, or a colleague answering the same card, changes nothing and sends nothing new.
The guest hears back, and it sticks
Your answer is relayed into the guest's chat in the concierge's voice. Choose to save it and it becomes property knowledge, so the same question never has to be escalated again.
Held questions in your dashboard, in priority order
The queue is ordered for you, and every open item is listed, oldest first.
Each alert carries enough to act without digging:
Questions and requests take different lanes, and a request is registered before the AI confirms it:
Styled after the dashboard's answer panel; the guest and property are invented.
What happens when you answer a held question
You type it once, in the dashboard. The system does the rest, and it is honest about the order things happen in.
Relayed to the guest in the concierge's voice
The concierge composes a plain, first-person message from what you said, in the same conversation where they asked. Guests hear one voice, and your team stays out of 2 AM threads.
The question leaves your inbox
The question is marked answered and leaves your inbox. If the guest asks the same open question again while they wait, it is not registered twice.
Saved as property knowledge, if you choose
With the save option ticked, your answer becomes team-written knowledge: scoped to the property, filed under a topic, and date-stamped as of today. Automatic learning may never overwrite it. It sits in your knowledge base with everything else the AI knows, and how the AI learns covers how saved answers add up over time.
Guest requests are closed out automatically
When your team marks an operational request done, or declines one, the guest is told automatically, in the same first-person voice. Nobody has to remember to circle back to the thread.
Two promises hold the whole flow together: the AI never promises a response time on your behalf, and it never tells a guest the team was asked unless the question was actually registered first.
Where escalation ends and stopping begins
Not everything the AI cannot answer becomes a held question. One kind of message ends the AI's turn in that conversation instead.
A question waits for you
It is held in the guest's own words, your team is told at once, and your answer is relayed back into the same conversation.
A complaint stops the AI
A complaint or a request for money back is not a held question. It stops the AI in that conversation entirely, and a person on your team is alerted.
The full set of guardrails is on when the AI stops. For how held questions fit into a whole team's day, see how TravelPal works for property management companies.
Where this goes next:
TravelPal is an AI guest messaging and upsell tool for vacation rental property managers, connecting to 33 property management systems including Guesty and Hostaway, priced at $25 per unit per month. Escalation is part of that flat rate, not an add-on.
Updated