The AI Agent can answer questions about a guest's own reservation — arrival and departure dates, how many nights, party size, booked extras, the total, and the booking status — but only after the conversation has been verified for that specific reservation.
Verification is what keeps this safe. Until it happens, the AI Agent has no access to any reservation data at all. Once it happens, the AI Agent answers the guest's original question in the same reply.
There is nothing to switch on. If your AI Agent is active, reservation answers are active.
What the AI Agent can answer
Once a conversation is verified, the AI Agent can use these details from the reservation:
Booking reference. To confirm which reservation the conversation is about.
Status. Expected, Checked In, Cancelled, and so on — the same labels your team sees in the inbox.
Arrival and departure. Actual check-in and check-out times where available, otherwise the booked stay dates.
Nights. The length of the stay.
Adults and children. Party size.
Cancellation and reason. If the booking was cancelled.
Confirmed at. When the booking was confirmed.
Purpose. Business or leisure.
Total and currency. For price questions.
Extras. Booked add-ons with name, quantity, price, and validity period — for example "2x Breakfast, 15 EUR, 4–6 August".
This list is fixed. Nothing else from the reservation reaches the AI Agent.
The reservation's stored contact details are never included. Phone numbers and email addresses on the booking are used to run the identity match and are never shown to the AI Agent, so they cannot appear in a reply.
What the guest has to do
One thing: give their booking reference.
When a guest asks a reservation question, the AI Agent asks for the booking reference. That is the only piece of information it is allowed to ask for. It then matches the contact the guest is writing from — their phone number on WhatsApp and SMS, their email address on email — against every contact stored on that reservation, including all occupants, not just the lead guest.
If it matches, the conversation is verified and the AI Agent answers the original question straight away.
The AI Agent never asks for a phone number or email address. Identity always comes from the channel itself, which the guest cannot influence. This is what makes the flow resistant to impersonation: there is no contact field for anyone to manipulate, so a guest cannot point verification at someone else's booking.
Which channels can verify
Verification depends on the channel being able to prove who the sender is.
Can verify:
WhatsApp — the sender's phone number is proven by the channel.
SMS — the sender's phone number is proven by the channel.
Email — the sender's address is proven by the channel.
Cannot verify:
Expedia and other OTA channels
Web messenger
Instagram, Telegram, and other social channels
On these channels the sender's real identity cannot be established, so the AI Agent will not verify. Instead, it asks the guest to write from WhatsApp, SMS, or email using the contact they booked with. If the guest cannot switch channel, the conversation is handed to a human.
This only affects reservation questions. On Expedia, the AI Agent continues to answer everything else exactly as before.
The Verified tag
Once a conversation is verified, a Verified tag appears next to the guest's name in the thread sidebar in your Unified Inbox.
This is the quickest way for your team to tell, at a glance, whether the AI Agent has reservation data available in that thread. If a guest is asking reservation questions and there is no Verified tag, verification has not happened or has expired.
How long verification lasts
Seven days. After a successful verification, every message on that channel carries the reservation context automatically for seven days. The guest is not asked again.
It belongs to the channel, not the person. The verification is held by the phone number or email address that proved itself. On a reservation with several occupants, the details stay available to whichever contact verified.
One reservation at a time. If a guest verifies a second booking on the same channel, the AI Agent works from the most recently verified one.
After seven days, the guest simply provides their booking reference again.
Guests replying to a campaign are verified automatically
If a guest replies to a reservation-triggered campaign — a pre-arrival message, for example — the conversation is verified before the AI Agent even runs. The guest never has to give a booking reference.
The logic is simple: your property already chose to send that specific contact a message about that specific reservation, so the reply itself proves who is writing.
What happens when verification fails
A wrong booking reference, someone else's booking reference, and a contact that is not on the reservation all produce the same neutral response: the AI Agent says it could not find the booking and asks the guest to double-check.
This is deliberate. If the AI Agent said "that reference exists but your number doesn't match", it would confirm the existence of other guests' bookings to anyone who guessed a reference. The identical response means the flow cannot be used to probe your reservations.
After one failed retry, the AI Agent hands the conversation to a human. Guests are never left stuck. If the sender's contact cannot be read at all, it hands off immediately.
A guest reporting "it says it can't find my booking" is usually one of the reasons in Common problems below, not a fault.
Testing it before guests use it
Admins can play through the whole flow in the Preview panel. Go to Inbox Settings > Bots, open your AI Agent, and use the Testing tab and Preview panel as described in Testing Your AI Agent with the Preview Panel.
The testing panel includes a channel selector, where you set the channel type and the contact the simulated guest is writing from. That lets you run through:
A trusted channel with a contact that is on the reservation — verification succeeds.
A wrong booking reference — the neutral failure response.
A contact that is not on the reservation — verification fails.
An untrusted channel — the AI Agent redirects instead of verifying.
The Preview panel runs the real verification logic but keeps everything in memory. Nothing is saved and no real guest data is touched. Verification carries across turns within one simulated conversation, and switching to a different thread or guest resets it.
Testing from your own phone will not work as a test of this feature. A staff number is not stored on any reservation, so verification will correctly fail. Use the Preview panel, or a real test reservation that has your contact details on it.
Common problems
If a guest cannot verify, work through these:
They are on a channel that cannot verify. Check the channel. Expedia, web messenger, and social channels cannot verify.
They booked with a different contact. The match runs against the contact details stored on the reservation. A guest who has since changed phone number or is writing from a different email address cannot verify from the new one. Ask them to write from the contact they booked with, or handle the request with a human.
The verification has expired. Grants last seven days. The guest just provides their booking reference again.
The booking reference is wrong. Guests often quote an OTA reference rather than the property's booking reference.
It is an older email conversation. Depending on how the conversation was created, some older email threads do not carry the sender's address in a way the system can check. These hand off to a human instead of verifying. This is a known gap and is being worked on.
Current limitations
No alternative contact flow yet. A guest who cannot verify from their booking contact has no self-service route. A flow that sends a code to another contact on the reservation is designed and may come later.
Group bookings are not supported. The current design handles one reservation per verification.
Verification in place on web messenger is not available. A code-based verification design exists and may return, which would let web messenger verify instead of redirecting.
Extras can occasionally show as unknown. Extras are fetched in a separate step. If that step fails, the AI Agent tells the guest it does not know the extras rather than guessing. Verification itself still succeeds and everything else is answered normally.
A note on cancelled bookings
Cancelled reservations verify normally. A guest can verify a cancelled booking and the AI Agent will tell them its status, including that it was cancelled and why.
Need Help?
Please contact us through the 'Talk to Us' option on the left menu in the platform, or through the Bookboost Support email at support@bookboost.io if you have questions or need additional support.