Guest message
Received Β· Jun 14, 2026, 4:26 p.m.
What is the upgrade?
Expedia Β· Message
Received Β· Jun 14, 2026, 4:26 p.m.
What is the upgrade?
Sent Β· Jun 14, 2026, 4:27 p.m.
Thank you for your interest in upgrading your room. We'll have one of our team members reach out to you to discuss the available options and pricing. Please allow us some time to review your request, and we'll be in touch soon.
Source: 1
Score Β· 54.15%Trigger Phrases
LLM Response β What to Send the Guest
| "We can guarantee connecting rooms in advance for an additional fee of $20 per room, per night. To confirm this for your stay, please provide the name and confirmation number for each reservation you'd like connected and we'll get this set up for you." |
|---|
Escalation Decision Rule
Trigger Phrases
LLM Response β What to Send the Guest
| "We'd be happy to look into upgrade and extension options for you. To have our hotel team reach out directly with availability and pricing, could you please share the best phone number to reach you?" |
|---|
| Note on Upgrade Information: Always set the expectation that upgrades are subject to availability. Do not confirm inventory or guarantee any specific room type. The human Concierge team handles all actual availability checks and pricing. |
|---|
Escalation Decision Rule
Source: 2
Score Β· 48.59%Upsell inquiries are initiated by the guest. The LLM presents the offer and collects guest intent. Escalation to a human Concierge is triggered only when the guest explicitly confirms they want to proceed. Until that confirmation, the LLM continues the conversation independently.
| Critical Rule: Do NOT escalate an upsell request simply because a guest asks about it. Only escalate after confirmed guest intent to purchase/proceed. Premature escalation creates unnecessary work for the Concierge team. |
|---|
Trigger Phrases
LLM Response β What to Send the Guest
| "Early check-in is available for $60 + taxes starting at 11:00 AM on your arrival date. Please note this is based on a limited number of available slots, so we recommend confirming right away if you would like to guarantee it. Would you like us to set this up for you?" |
|---|
Escalation Decision Rule
Trigger Phrases
LLM Response β What to Send the Guest
| "We'd be happy to arrange a guaranteed late check-out for you. Here are the available options:" |
|---|
| Checkout Time | Tue β Fri Departure | Sat β Mon Departure |
|---|---|---|
| 12:00 PM | $20 + tax | $30 + tax |
| 1:00 PM | $30 + tax | $40 + tax |
| 2:00 PM | $40 + tax | $50 + tax |
| 3:00 PM | $50 + tax | $60 + tax |
| 4:00 PM | $60 + tax | $70 + tax |
| "To secure your preferred departure time in advance, just let us know which time works best and we'll get that arranged. If you'd prefer not to pay a fee, late checkout is also available based on availability by calling the front desk directly on the morning of departure." |
|---|
Escalation Decision Rule
Trigger Phrases
LLM Response β What to Send the Guest
Source: 3
Score Β· 45.92%RAG Context: Section 1 upsell escalation rules for Paris Las Vegas β guest-initiated upsell flows for Early Check-In Guarantee, Late Check-Out Guarantee, Connecting Rooms, and Room Upgrade / Stay Extension. For each: trigger phrases, the exact message to send the guest, and the escalation decision rule. Core rule: do NOT escalate simply because a guest asks; escalate only after confirmed guest intent to purchase/proceed, and send a holding reply first. Part of the master framework in
paris-las-vegas-resort-casino-escalations.md.
Upsell inquiries are initiated by the guest. The LLM presents the offer and collects guest intent. Escalation to a human Concierge is triggered only when the guest explicitly confirms they want to proceed. Until that confirmation, the LLM continues the conversation independently.
This response was escalated. Was that decision correct?
Not yet assessed
Set by the preferred response below β saving a correction marks it incorrect, clearing the correction marks it correct.
Not yet assessed
Optionally provide a better response. Saving non-empty text marks the LLM response incorrect; clearing it marks the LLM response correct.