Guest message
Received Β· Jun 17, 2026, 9:06 p.m.
Good evening, Hope that you are doing well. I send this message as a follow up on the status of the deposit refund for both reservations. Thanks and kind regards Carlos
Expedia Β· Message
Received Β· Jun 17, 2026, 9:06 p.m.
Good evening, Hope that you are doing well. I send this message as a follow up on the status of the deposit refund for both reservations. Thanks and kind regards Carlos
Sent Β· Jun 17, 2026, 9:07 p.m.
Thank you for reaching out to us again, Carlos. We appreciate your patience. Your deposit refund is currently being processed.
Source: 1
Score Β· 59.12%RAG Context: Escalation guideline for Paris Las Vegas Resort Casino guest messages where the guest (a) requests a reply by email ("can you email me?", "reply to my email", "what's your email?", "I'll send you my email") or (b) sends an email address β including a platform-REDACTED token that looks like an email, e.g.
----@--------.---,****@*****.***,xxxx@xxxxx.xxx(third-party booking channels such as Expedia / Booking.com auto-mask email addresses in guest messages, so a redacted-email token signals the guest tried to share their email). The automated assistant operates on the booking channel's text-messaging and post-stay review threads and cannot send or receive email, and because the reservation was made through a third-party provider the property has no direct email relationship with the guest β so email communication is not possible. The assistant sends the standard reply stating this (booking made through a third-party provider) and escalates to a human Concierge. Maps to tag EMAIL-CONTACT; part of the master framework inparis-las-vegas-resort-casino-escalations.md. A guest simply asking for the property's general contact info for a routine reason is handled fromparis-las-vegas-resort-casino-property-details.md. Always escalate on trigger; do not email the guest and do not echo a full real email address back in a public reply.
Source: 2
Score Β· 58.04%Escalation decision rules for automated guest messaging and post-stay reviews.
This framework trains the LLM automated concierge system on exactly two things:
| Core operating principle: The LLM handles routine information requests and upsell conversations up to β but not including β the moment a guest commits or a sensitive situation arises. At that boundary, the LLM sends the guest a holding reply and immediately generates an escalation email to the human Concierge team. |
|---|
paris-las-vegas-resort-casino-upsell-escalations.md (Early Check-In, Late Check-Out, Connecting Rooms, Room Upgrade / Stay Extension). Escalate only after confirmed guest intent to purchase/proceed.paris-las-vegas-resort-casino-general-escalations.md (Name Additions, Stay Modifications, Invoice/Folio, Room Issues, Maintenance, Noise/Safety, Billing, General Complaints).paris-las-vegas-resort-casino-review-response-escalations.md (positive and negative/mixed review handling and escalation triggers).paris-las-vegas-resort-casino-confirm-fee-escalations.md (CONFIRM-FEE) β catch-all when a guest accepts or authorizes any fee or surcharge; backs the Section 1 upsell rules.Use this table as a fast lookup for any incoming guest message. Find the request type, confirm whether the LLM handles it or escalates, and identify the specific trigger that activates the escalation.
Source: 3
Score Β· 55.67%Use this table as a fast lookup for any incoming guest message. Find the request type, confirm whether the LLM handles it or escalates, and identify the specific trigger that activates the escalation.
| Request / Scenario | LLM Response | Escalation Trigger |
|---|---|---|
| Early Check-In | Share $60 + tax fee, explain availability | Guest confirms they want to pay β then escalate |
| Late Check-Out | Share fee schedule, explain availability option | Guest confirms fee + desired time β then escalate |
| Connecting Rooms | Explain $20/room/night fee, request names & conf #s | Guest confirms fee + provides all names & conf #s β then escalate |
| Room Upgrade / Extend Stay (incl. adding a night or a later checkout) | Ask for best contact phone number; do not route to the booking channel | Guest provides phone number β then escalate |
| Fee / Charge Confirmation (any) | Thank guest, acknowledge the accepted fee; do not apply the charge or guarantee the add-on | Guest accepts/authorizes any fee or surcharge β always escalate (CONFIRM-FEE) |
| Name Addition / Alt Check-In | Collect name, explain ID/card requirements | Primary name change, billing auth, identity mismatch |
| Stay Modification / Cancel (shorten, cancel, date shift, room change, primary-name change β not adding nights or an additional guest name) | Explain 3rd-party policy, direct to booking channel | 3rd party cannot process; complex booking changes |
| Invoice / Folio / Receipt (incl. itemized) | Send the self-service folio response (emailed after checkout; check spam; else caesars.com folio-request link) | Billing dispute / correction / refund; business-invoice format; incidentals hold past timeframe |
| Room Issue / Housekeeping | Apologize, gather details, direct to front desk first | Front desk unresponsive; safety impact; comp requested |
| Maintenance Request | Acknowledge, direct to front desk / Laurel text | Front desk unresponsive; repair still needed |
| Noise / Safety Complaint | Acknowledge, direct to Security / front desk | Security/front desk intervention needed; unresolved |
| Billing Inquiry / Dispute / Refund | Acknowledge, collect details | Any correction, dispute, or refund β always escalate |
| General Complaint / Comp | Empathize, document thoroughly | Compensation requested; multiple failures; emotional guest |
| Post-Stay Review β Negative | Acknowledge publicly with empathy template | Safety, discrimination, viral-risk β escalate within 12h |
| Post-Stay Review β Positive | Respond with standard thank-you template | No escalation required |
This response was not escalated. Was that decision correct?
Marked correct
Set by the preferred response below β saving a correction marks it incorrect, clearing the correction marks it correct.
Marked incorrect
Optionally provide a better response. Saving non-empty text marks the LLM response incorrect; clearing it marks the LLM response correct.