What happens when a guest asks for a refund?
TravelPal's AI stops. The reply it drafted is withheld, the conversation is flagged for a person, and your team is notified by a team chat message and an email that the AI has gone quiet in that thread.
Every inbound guest message is checked for whether the guest is seeking money back or filing a complaint: a refund ask, a request for compensation, a formal platform resolution case, or a threat to escalate or leave a bad review over something gone wrong.
On a hit, the AI sends nothing. The whole conversation is marked human-only, so the AI writes no further messages in that thread. A polite follow-up from the guest, like an ok thanks, does not restart it; the thread stays with your team until a person takes it over. The AI never negotiates a refund, never offers compensation, and never argues a complaint. Those conversations are handed to a person by design.
Your team hears about it twice
The moment the AI steps back, your team is told two ways at once: a message in your team channel and an email to the team inbox, both carrying the guest's message and saying the AI has stopped replying. When a person steps in, the full thread is already in the inbox of your property management system, Guesty or whichever system you run, because every message the AI sends goes out through that system rather than around it.
Ordinary problems are not complaints
A guest saying the wifi is slow or nobody brought towels is asking you to fix something, not filing a claim, and the AI deliberately keeps helping with those instead of going silent. The line it watches for is money and formal escalation, not frustration. See what happens when the AI does not know an answer for the other way conversations reach your team, or book a demo to watch the handoff live.