πŸ’¬

Expedia Β· Message

Flamingo Las Vegas Hotel & Casino

Autorag source
flamingo-las-vegas-hotel-casino
Thread ID
f8071210-4390-4d24-81b6-a9e4faa84914
Message ID
9d5ddc63-c4de-4a95-9e91-33c1dabe2f94
Event ID
741fb128-c18c-4a41-97f7-a51dee0bc9a8
🧳

Guest message

Received Β· Jun 16, 2026, 10:12 p.m.

stay-modification

I can give you there names and check in dates

🏨

Property Response

Sent Β· Jun 16, 2026, 10:13 p.m.

stay-modification-policy

Thank you for reaching out. For any changes to your reservation, including adding or modifying rooms, please contact Expedia customer service directly. They will be able to assist you with your request.

Escalation Information
Possible escalation
Yes
Actual escalation
No
Possible escalation reason
RAG retrieval returned escalation file "flamingo-las-vegas-hotel-casino/flamingo-las-vegas-hotel-casino-customer-phone-number-escalations.md" scoring 58% (at or above the 45% escalation threshold).
Actual escalation reason
Escalation judge LLM did not indicate escalation: The guest's message does not contain a phone number or a request to be called or texted, so the customer-phone-number escalation rule does not apply.
RAG Context Documents
  1. Source: 1

    Score Β· 57.77%

    flamingo-las-vegas-hotel-casino/flamingo-las-vegas-hotel-casino-customer-phone-number-escalations.md

    Show source text

    flamingo customer-phone-number

    RAG Context: Escalation guideline for Flamingo Las Vegas guest messages where the guest's message contains a phone number β€” i.e. the guest supplies a phone number (their own cell / mobile, a callback number, an alternate contact, or someone else's) or asks to be called or texted at a number (e.g. "Please call me at (702) 555-0199."). The Flamingo automated assistant operates on text-messaging and post-stay review channels and cannot place or receive phone calls or SMS; a guest sharing a personal phone number is requesting β€” explicitly or implicitly β€” direct human or voice contact, and is supplying personal contact information (PII) that a person must handle. Every guest message that contains a phone number is therefore an immediate, always-on escalation to a human Concierge. Maps to risk category CUSTOMER PHONE NUMBER (tag CUSTOMER-PHONE-NUMBER) in the master Flamingo Risk Management Framework (flamingo-las-vegas-hotel-casino-escalations.md). All matches require immediate escalation. Do not call or text the number, and do not repeat the full number back in a public reply β€” acknowledge and route to a human.

    Customer-Supplied Phone Number

    Guest asks / says:

    • "Please call me at (702) 555-0199."
    • "My cell is 702-555-0142 β€” call anytime."
    • "Can someone reach me at +1 555 867 5309?"
    • "Text me at 5550100123."
    • "Here's my number: 555-0188, please follow up by phone."
    • "Call my husband at 702.555.0177 to sort this out."
    • "I'd rather talk β€” my number's 555 0150."

    Trigger keywords (non-exhaustive): call me at, reach me at, text me at, my number is, my cell is, my mobile is, here's my number, you can reach me on, call my [husband / wife / assistant] at, call me back, give me a call β€” plus any digit sequence formatted as a phone number in the guest's message: (NPA) NXX-XXXX, NPA-NXX-XXXX, NPA.NXX.XXXX, 7-digit NXX-XXXX, international +CC … / +1 NPA NXX XXXX, or a bare 10-/11-digit run.

    Disambiguation: This category fires when the guest supplies or shares a phone number (or asks to be contacted at one). A guest asking for the property's phone number ("what's the front-desk / pool / spa number?") is a routine information request β€” answer from flamingo-las-vegas-hotel-casino-property-details.md, do not tag here. Do not over-trigger on non-phone digit strings: reservation / confirmation numbers, Caesars Rewards numbers, room numbers, dates, and dollar amounts are not phone numbers. Tag only when the digits are presented as a phone / contact number or paired with a call/text request. When a message both shares a number and raises another issue (billing, safety, complaint), apply both tags.

    TAG as customer-phone-number

    Respond with

    Thank the guest and let them know a team member will follow up. Send a holding reply only (e.g. "Thank you β€” I've passed your details to our team and someone will be in touch."). Never state that a call or text has been placed, returned, or scheduled for a specific time; do not call or text the number yourself; and do not echo or repeat the full phone number back in the reply β€” especially on a public review channel β€” treat it as private contact information.

    Before escalating

    Do not call or text the guest and do not promise a specific callback time. The LLM is authorized to acknowledge only. Capture in the escalation record: the guest's exact wording, the phone number provided and whose number it is (guest / spouse / assistant), any preferred call or text window, the reason they want to be contacted, the inbound channel, and full reservation details β€” so the Concierge can make voice/SMS contact.

    Escalate when:

    • Always. Any guest message that contains a phone number, or any request to be called or texted, is an immediate escalation so a human can make voice/SMS contact and handle the personal contact details β€” regardless of the reason, tone, or topic. The LLM has no ability to place or receive calls or texts.
    • If the phone number accompanies a sensitive issue (safety, injury, theft, health, harassment, legal, billing dispute), escalate under both this tag and the relevant risk-category guide, and flag the higher-severity tier.
    • Tag the conversation: [CUSTOMER-PHONE-NUMBER] + [TIER LEVEL] + [CHANNEL: Expedia] and route to the Flamingo Concierge queue with priority flag.
  2. Source: 2

    Score Β· 56.55%

    flamingo-las-vegas-hotel-casino/flamingo-las-vegas-hotel-casino-general-escalations.md

    Show source text

    RAG Context: Section 2 general escalation rules for Flamingo Las Vegas routine guest-service requests β€” Name Additions & Guest Authorization, Stay Modifications & Cancellations, Invoice/Folio/Receipt Requests, Room Issues & Housekeeping, Maintenance Requests, Noise/Neighbor/Safety Complaints, Billing Issues & Charge Disputes, and General Complaints & Compensation Requests. For each: trigger phrases and a clear split between what the LLM responds to directly versus what it escalates to a human Concierge. Safety, medical, injury, theft, and billing-dispute messages always escalate. The on-site Security Team is available 24/7 at 702-733-3523. Part of the master framework in flamingo-las-vegas-hotel-casino-escalations.md.

    Section 2 β€” General Escalation Rules

    These rules apply to routine guest-service requests that arrive via message. The LLM should acknowledge professionally, provide any information it can, and escalate when the situation requires human action or decision-making.

    E β€” Name Additions & Guest Authorization

    Trigger Phrases

    • "Can you add a name to my reservation?"
    • "Can someone else check in before I arrive?"
    • "Can I change the name on my booking?"
  3. Source: 3

    Score Β· 55.81%

    flamingo-las-vegas-hotel-casino/flamingo-las-vegas-hotel-casino-conduct-harassment-escalations.md

    Show source text

    flamingo conduct-harassment

    RAG Context: Escalation guideline for Flamingo Las Vegas guest reports involving sexual harassment by staff or another guest, discriminatory treatment (race, gender, disability, religion, etc.), verbal abuse or threatening behavior from staff, profiling/surveillance/targeting, hostile guest-to-guest interactions, complaints about use of force by security personnel, or ADA accommodations denied or mishandled. Maps to risk category HARASSMENT & CONDUCT (tag CONDUCT-HR) in the master Flamingo Risk Management Framework (flamingo-escalations.md). All matches require immediate escalation. The LLM must not investigate, validate, or dismiss the claim.

    E. Harassment & conduct

    Guest asks / says:

    • "I was sexually harassed by a staff member or another guest."
    • "I experienced discrimination based on race / gender / disability / religion / orientation."
    • "A staff member was verbally abusive / threatening to me."
    • "I felt profiled / surveilled / targeted on the property."
    • "There was a hostile or aggressive interaction with another guest."
    • "I have a complaint about use of force by security personnel."
    • "My ADA accommodation was denied or mishandled."

    Trigger keywords (non-exhaustive): harassed, sexual harassment, touched me, inappropriate, discriminated, racist, sexist, bias, profiled, staff was aggressive, threatening, yelled at me, abusive, felt unsafe, threatened, intimidated, followed, ADA, disability, accessibility, wheelchair, accommodation denied.

    TAG as conduct-harassment

    Respond with

    Use one of the approved Risk Management response templates from flamingo-escalations.md Β§5. Acknowledge with grounded seriousness. Do not investigate, validate, dismiss, name staff or other guests, describe internal HR or security protocols, or hint at compensation. Avoid any language that could read as taking sides.

    Before escalating

    Do not investigate or attempt to resolve. The LLM is authorized to acknowledge and empathize only. Capture the guest's exact wording in the escalation record β€” wording matters in HR and conduct matters. The concierge will route to property leadership and HR as appropriate.

    Escalate when:

    • Always. Any mention of sexual harassment, discrimination, verbal abuse, profiling, threatening behavior, security use-of-force, or ADA accommodation failure is an immediate escalation regardless of star rating, tone, or whether the guest demands action.
    • Tag the conversation: [CONDUCT-HR] + [TIER LEVEL] + [CHANNEL: Expedia] and route to the Flamingo Concierge queue with priority flag.

Escalation decision review

This response was not escalated. Was that decision correct?

Marked correct

LLM response review

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.