[
  {
    "id": "guest-message-triage",
    "title": "Triage guest messages before the front desk replies",
    "description": "Turn incoming requests into draft answers, service tasks, and a short escalation queue.",
    "category": "Guest service",
    "tool": "Approved AI assistant",
    "toolUrl": null,
    "level": "Beginner",
    "evidence": "Vendor example",
    "output": "A review queue with intent, urgency, source policy, draft response, and assigned department.",
    "inputs": [
      "A small set of redacted guest requests",
      "Current property FAQ and department escalation rules"
    ],
    "connections": "Begin with approved exports or pasted source material. Live access requires separately configured integrations and permissions; this guide does not install them.",
    "prompt": "Review these guest requests for [property] using only [approved FAQ]. For each, return intent, urgency, relevant policy reference, suggested department, and a draft response. Mark missing facts as unknown. Escalate safety issues, complaints, refunds, and policy exceptions to staff. Do not send messages or create service commitments. Keep the guest's actual request beside your interpretation.",
    "steps": [
      "Choose twenty historical requests and label the correct action with the front-desk lead.",
      "Provide the current FAQ with version dates and an explicit escalation list.",
      "Run the sample and compare answers and routing with the staff labels.",
      "Allow drafts into a staff queue only after correcting recurring errors."
    ],
    "review": "Front-desk staff check policy accuracy, tone, urgency, and room-specific facts before replying.",
    "limits": [
      "A plausible answer can still quote an outdated policy.",
      "Languages, PMS access, and live ticket creation depend on the implementation."
    ],
    "measure": "Missed escalations, factual corrections per draft, and staff review time.",
    "sources": [
      {
        "title": "AI guest messaging",
        "url": "https://us.canarytechnologies.com/products/ai-guest-messaging",
        "author": "Canary Technologies",
        "note": "Vendor describes guest answers, translation, service tickets, and staff handoff. Our prompt and review process are editorial additions.",
        "kind": "Vendor example"
      },
      {
        "title": "Booking.com partner agent for guest inquiries — X post",
        "url": "https://x.com/victorialslocum/status/1993636038313443826",
        "author": "Victoria Slocum",
        "note": "The visible X post reports an agent helping Booking.com accommodation partners respond to guest inquiries. The public embed truncates further details; it does not establish routing features, accuracy, or savings. The prompt and escalation checks here are our editorial additions.",
        "kind": "Practitioner report"
      }
    ],
    "reviewed": "2026-10-12"
  },
  {
    "id": "phone-intake-handoff",
    "title": "Prepare a complete handoff from a hotel phone inquiry",
    "description": "Capture the reason for the call and the next action so staff do not start over.",
    "category": "Guest service",
    "tool": "Approved AI assistant",
    "toolUrl": null,
    "level": "Intermediate",
    "evidence": "Vendor example",
    "output": "A call-intake summary and a tested transfer checklist.",
    "inputs": [
      "Approved call transcripts or staff notes",
      "Property phone menu and escalation contacts"
    ],
    "connections": "Begin with approved exports or pasted source material. Live access requires separately configured integrations and permissions; this guide does not install them.",
    "prompt": "For each [approved call transcript], summarize the caller's stated request, stay dates if supplied, property, unanswered questions, and best staff destination. Separate stated facts from guesses. Draft a one-paragraph handoff for the receiving employee. Flag urgent or sensitive matters. Do not promise availability, payment, refunds, or a completed booking.",
    "steps": [
      "Use approved transcripts to define the fields staff need during a transfer.",
      "Test summaries against a small sample of common and unusual inquiries.",
      "Confirm that staff destinations and after-hours coverage are current.",
      "If connecting a voice product, test transfer failures and a staff fallback before launch."
    ],
    "review": "A staff owner verifies that the caller can reach a human and that no important context is lost.",
    "limits": [
      "Live calling requires a configured voice provider and appropriate recording procedures.",
      "A transcript can omit tone, background noise, or details the caller corrected."
    ],
    "measure": "Handoff completeness, failed transfers, and repeated questions after transfer.",
    "sources": [
      {
        "title": "Hotel Zephyr: AI voice intake",
        "url": "https://us.canarytechnologies.com/customers/hotel-zephyr",
        "author": "Canary Technologies / Hotel Zephyr",
        "note": "Customer story describes call handling and contextual transfer to staff. Its results are vendor-reported, not an independent benchmark.",
        "kind": "Vendor example"
      }
    ]
  },
  {
    "id": "pre-arrival-brief",
    "title": "Draft a useful pre-arrival message",
    "description": "Give guests the information they need before check-in, with approved links to available services.",
    "category": "Reservations",
    "tool": "Approved AI assistant",
    "toolUrl": null,
    "level": "Beginner",
    "evidence": "Vendor example",
    "output": "A pre-arrival draft with arrival instructions, verified service links, and unresolved questions.",
    "inputs": [
      "Approved arrival instructions and service catalog",
      "Guest stay details appropriate for this message"
    ],
    "connections": "Begin with approved exports or pasted source material. Live access requires separately configured integrations and permissions; this guide does not install them.",
    "prompt": "Draft a pre-arrival message for [property] and [stay details]. Use [approved instructions] for check-in, parking, access, and contact details. Include only relevant services from [catalog] with their official links. Distinguish confirmed details from questions for the guest. Do not imply a reservation or paid add-on is confirmed. Keep the message concise and return a factual verification checklist.",
    "steps": [
      "Collect the latest arrival details and links maintained by each department.",
      "Define which information is appropriate for each guest segment.",
      "Generate drafts for a few representative stays and review them on mobile.",
      "Schedule approved messages only after checking contact preferences and current availability."
    ],
    "review": "Reservations staff verify dates, accessibility details, links, contact permission, and service availability.",
    "limits": [
      "A service catalog is not live inventory.",
      "Incorrect access instructions can disrupt arrival; always retain a staff contact."
    ],
    "measure": "Guest clarification requests, broken links, and corrections before sending.",
    "sources": [
      {
        "title": "La Butte: pre-arrival messaging",
        "url": "https://www.hijiffy.com/success-stories/la-butte",
        "author": "HiJiffy",
        "note": "Case study describes pre-arrival WhatsApp information and links to hotel services. Our guide does not promise the same results.",
        "kind": "Vendor example"
      }
    ]
  },
  {
    "id": "after-hours-booking-queue",
    "title": "Qualify an after-hours booking inquiry",
    "description": "Collect the missing details and prepare a clean handoff to reservations.",
    "category": "Reservations",
    "tool": "Approved AI assistant",
    "toolUrl": null,
    "level": "Beginner",
    "evidence": "Vendor example",
    "output": "A structured inquiry containing dates, party size, property preference, and a next-action draft.",
    "inputs": [
      "Booking inquiries from approved channels",
      "Booking-engine links and reservation policies"
    ],
    "connections": "Begin with approved exports or pasted source material. Live access requires separately configured integrations and permissions; this guide does not install them.",
    "prompt": "Read [inquiries] for [hotel group]. Extract requested dates, adults and children, room requirements, property preference, and stated special requests. Mark absent values unknown and suggest the fewest questions needed to proceed. Provide an official booking link where relevant. Do not invent rates, room inventory, or a reservation confirmation. Produce a staff review queue and an unsent reply draft.",
    "steps": [
      "Define the fields reservations needs to handle an inquiry.",
      "Use historic examples to check date ambiguity and party-size extraction.",
      "Add read-only booking access only if its permissions and failure behavior are clear.",
      "Route unresolved or policy-sensitive cases to reservations on opening."
    ],
    "review": "Check rates and availability in the booking system immediately before giving a quote.",
    "limits": [
      "A drafted reply is not a booking.",
      "Date formats, time zones, and special requirements need explicit checks."
    ],
    "measure": "Complete inquiries received, incorrect field extractions, and unresolved handoffs.",
    "sources": [
      {
        "title": "The Address Collective case study",
        "url": "https://asksuite.com/blog/the-address-collective-case/",
        "author": "Helena Neves / Asksuite",
        "note": "Vendor describes digital booking conversations and lead qualification outside office hours. We do not use its ROI claim as a forecast.",
        "kind": "Vendor example"
      }
    ]
  },
  {
    "id": "complaint-action-queue",
    "title": "Turn guest complaints into a department action queue",
    "description": "Group recurring issues while keeping urgent individual cases visible.",
    "category": "Property operations",
    "tool": "Approved AI assistant",
    "toolUrl": null,
    "level": "Beginner",
    "evidence": "Editorial adaptation",
    "output": "An issue register with supporting guest excerpts, department owners, and proposed next steps.",
    "inputs": [
      "Redacted complaints or guest-review exports",
      "Department ownership and urgency rules"
    ],
    "connections": "Begin with approved exports or pasted source material. Live access requires separately configured integrations and permissions; this guide does not install them.",
    "prompt": "Analyze [guest feedback] for [property]. Group specific operational issues and preserve a source reference for each. Show frequency within this sample, contradictory feedback, and urgent individual cases separately. Suggest a department owner and a next investigation step. Do not infer guest demographics or decide compensation. Draft no public response unless requested; label any response as requiring staff approval.",
    "steps": [
      "Assemble a bounded feedback sample with dates and source IDs.",
      "Have department heads define urgent issues that must never be hidden in a cluster.",
      "Review the first issue register against the original messages.",
      "Assign approved actions to owners and record the resolution outside the model."
    ],
    "review": "Duty managers review safety issues immediately; department heads check that categories reflect the actual complaint.",
    "limits": [
      "Sentiment classification can misread irony or translation.",
      "Repeated comments may refer to one incident, not independent failures."
    ],
    "measure": "Missed urgent issues, useful actions accepted, and recurring unresolved themes.",
    "sources": [
      {
        "title": "How AI supports hotel teams",
        "url": "https://www.hijiffy.com/resources/articles/10-ways-hijiffy-ai-supports-hotel-teams",
        "author": "Lucie Maeder / HiJiffy",
        "note": "Vendor describes response suggestions, translation, sentiment detection, and routing. Human decisions remain part of this editorial workflow.",
        "kind": "Vendor example"
      }
    ]
  },
  {
    "id": "revenue-scenario-brief",
    "title": "Review pickup, pricing, and hotel revenue forecasts",
    "description": "Compare booking pace, cancellations, occupancy, ADR, and RevPAR before the revenue meeting.",
    "category": "Revenue management",
    "tool": "Approved AI assistant",
    "toolUrl": null,
    "level": "Intermediate",
    "evidence": "Editorial adaptation",
    "output": "A source-linked scenario table with assumptions and questions for the revenue manager.",
    "inputs": [
      "Approved occupancy, pickup, ADR, and cancellation exports",
      "Revenue manager assumptions and known event dates"
    ],
    "connections": "Begin with approved exports or pasted source material. Live access requires separately configured integrations and permissions; this guide does not install them.",
    "prompt": "Prepare a revenue review for [property] and [stay-date range] using [dated PMS and revenue exports]. Reconcile rooms sold, rooms available, cancellations, and room revenue. Define net pickup as the change in rooms on the books for the same stay dates between the supplied snapshots. Compare booking pace at the same lead time, with weekday and seasonal differences noted. Compute occupancy as rooms sold / rooms available, ADR as room revenue / rooms sold, and RevPAR as room revenue / rooms available; mark ratios with zero denominators as not calculable and use consistent inventory and revenue definitions. Show formulas and distinguish actuals, on-the-books values, and forecasts. Explain unusual dates using verified event or aviation evidence where available, without claiming causation. Compare only the supplied scenario assumptions, show uncertainty, and list questions on pricing, distribution, and stay restrictions. Do not change rates or invent forecast values.",
    "steps": [
      "Agree the decision and reporting period with the revenue manager.",
      "Reconcile inputs and define scenario assumptions outside the model.",
      "Generate the brief and recompute key totals in a spreadsheet.",
      "Discuss the scenarios and record approved decisions with an owner."
    ],
    "review": "Revenue staff verify formulas, inventory definitions, event context, and the difference between forecast and actuals.",
    "limits": [
      "Forecasts are uncertain and can miss local changes.",
      "Vendor forecasting capabilities are not recreated by a prompt alone."
    ],
    "measure": "Reconciliation errors, assumptions documented, and decisions traceable to inputs.",
    "sources": [
      {
        "title": "Cloudbeds Signals",
        "url": "https://signals.cloudbeds.com/",
        "author": "Cloudbeds",
        "note": "Product page describes hospitality forecasting and decision support. This guide proposes a manual scenario review, not autonomous rate changes.",
        "kind": "Vendor example"
      }
    ],
    "reviewed": "2026-10-12"
  },
  {
    "id": "guest-service-scheduling-handoff",
    "title": "Prepare a guest-service scheduling handoff",
    "description": "Turn a guest request into a complete scheduling brief before staff confirm an appointment or service.",
    "category": "Property operations",
    "tool": "Approved AI assistant",
    "toolUrl": null,
    "level": "Beginner",
    "evidence": "Editorial adaptation",
    "output": "A staff queue showing the requested service, preferred time, missing details, and the department that must confirm it.",
    "inputs": [
      "Redacted guest service requests and their source IDs",
      "Approved service hours, scheduling rules, and department contacts"
    ],
    "connections": "Start with approved message exports. Availability checks require a separately configured scheduling connection. This template does not reserve a slot or contact a guest.",
    "prompt": "Use [guest requests] and [property service rules] to prepare a scheduling handoff. For each request, preserve the source ID and exact requested date or time; extract the service, stated party size, relevant constraints, responsible department, and missing information. Distinguish a guest preference from confirmed availability. Flag ambiguous dates, time zones, and requests outside published service hours. Draft the shortest clarification question needed. Do not book, cancel, send a message, promise a slot, or invent a service price. Return the queue for staff approval.",
    "steps": [
      "Choose a small set of historical requests for services your property actually offers.",
      "Have the responsible department define required fields, service hours, and escalation rules.",
      "Run the prompt, comparing extracted dates and requirements with the original requests.",
      "Have staff check the scheduling system and confirm availability before sending any reply."
    ],
    "review": "The service owner checks dates, availability, special requirements, and guest permission before confirming a service.",
    "limits": [
      "The source describes a platform spanning housing and hospitality; it does not prove this exact hotel workflow was deployed.",
      "A text model cannot know live capacity from a service-hours document or a stale calendar export."
    ],
    "measure": "Missing scheduling details, incorrect date extractions, and back-and-forth messages needed before staff confirmation.",
    "reviewed": "2026-10-12",
    "sources": [
      {
        "title": "Conduit housing and hospitality agent platform — X announcement",
        "url": "https://x.com/ycombinator/status/1910347038660657592",
        "author": "Y Combinator",
        "note": "Y Combinator describes Conduit agents for leasing, onboarding, scheduling, and retention with human handoff. This service-scheduling checklist is an editorial adaptation of that broader announcement, not a tested Conduit setup or a claim of achieved savings.",
        "kind": "Vendor example"
      }
    ]
  },
  {
    "id": "concert-event-demand",
    "title": "Plan room demand around concerts and local events",
    "description": "Compare concerts, festivals, conventions, and sports dates with booking pace before reviewing rates and stay restrictions.",
    "category": "Revenue management",
    "tool": "AI assistant + verified data",
    "toolUrl": null,
    "level": "Intermediate",
    "evidence": "Editorial adaptation",
    "output": "A dated event calendar showing affected stay nights, booking pace, evidence, and decisions for the revenue meeting.",
    "inputs": [
      "Official event dates, venue locations, cancellation notices, and dated attendance estimates",
      "Property pickup, rooms on the books, available rooms, ADR, and comparable historical stay dates",
      "Current rates, stay restrictions, and the revenue manager’s approved scenario assumptions"
    ],
    "connections": "Start with a venue calendar and PMS exports. Event providers such as PredictHQ can supply a licensed feed. Public X posts can flag an announcement; confirm dates and status with the organizer before using it. Live feeds and PMS access need separate integration.",
    "prompt": "Build an event-demand brief for [property and location] over [stay-date range] from [dated event records] and [PMS exports]. Cite the source and retrieval time for every event. Deduplicate festival listings and identify cancellations or date changes. Map likely arrival and departure nights, including nights before and after the event. Compare rooms on the books and net pickup with matched historical dates at the same booking lead time. Separate observed bookings from assumptions. Treat attendance as an event estimate, not room demand. Show low/base/high room-night scenarios only when [approved conversion assumptions] are provided; otherwise leave impact unknown. List rate, minimum-stay, marketing, and staffing questions for review. Do not change rates or restrictions or claim an uplift was caused by the event.",
    "steps": [
      "Choose the property catchment and a 90-day stay window; verify event dates and travel times to the venue.",
      "Match prior events and ordinary dates by weekday, season, booking lead time, and available inventory.",
      "Join event records to pickup by stay date; record overlapping events and compare explicit scenarios in a spreadsheet.",
      "Review rates, stay restrictions, packages, and staffing with the revenue manager; record the decision and check again if the event changes.",
      "After the event, compare actual room nights and net room revenue with the baseline and record forecast error."
    ],
    "review": "The revenue manager confirms event status, comparable dates, inventory, and scenario math before approving any price or restriction change.",
    "limits": [
      "Attendance, venue capacity, and social engagement are not hotel bookings. Local audiences may need no overnight stay.",
      "A sold-out hotel can hide unmet demand; observed occupancy alone does not measure the full opportunity.",
      "Overlapping events, displaced regular guests, and cancellation risk make simple attribution unreliable."
    ],
    "measure": "Forecast error by stay date, net pickup, occupancy, ADR, RevPAR, and room revenue after channel costs; compare with the documented baseline.",
    "sources": [
      {
        "title": "Optimand: combining event data with hotel search and revenue signals",
        "url": "https://www.optimand.com/casestudies/predicthq/",
        "author": "Optimand",
        "note": "The provider describes combining local events, pickup, pricing, and hotel web searches. Our calendar, prompt, and decision checklist are editorial additions; no uplift is promised.",
        "kind": "Vendor example"
      },
      {
        "title": "Accommodation demand before and after events",
        "url": "https://docs.predicthq.com/getting-started/predicthq-data/impact-patterns",
        "author": "PredictHQ",
        "note": "Documentation describes leading and lagging event effects for accommodation. It supports considering surrounding stay nights, not a fixed uplift or guaranteed forecast.",
        "kind": "Vendor example"
      }
    ],
    "reviewed": "2026-10-12"
  },
  {
    "id": "flight-disruption-demand",
    "title": "Review hotel demand when flights are delayed or canceled",
    "description": "Separate possible overnight stays near the airport from missed arrivals and cancellations at your property.",
    "category": "Revenue management",
    "tool": "AI assistant + verified data",
    "toolUrl": null,
    "level": "Intermediate",
    "evidence": "Editorial adaptation",
    "output": "A short disruption brief with verified flight changes, exposed stays, room availability, and actions for reservations and revenue staff.",
    "inputs": [
      "Timestamped airport or airline status reports, or a licensed flight-status feed",
      "Approved arrival and cancellation summaries, pickup, room availability, and existing airline allotments",
      "Shuttle capacity, check-in rules, staffing coverage, and approved commercial terms"
    ],
    "connections": "Use airline or airport notices first; a service such as FlightAware AeroAPI can provide flight status and configured alerts. Join to approved hotel summaries. Flight data does not include a list of passengers who need your hotel.",
    "prompt": "Prepare a disruption brief for [property], [airport], and [local-time window] using [timestamped flight updates] and [hotel summaries]. Deduplicate flights by operating flight and departure date, separate codeshares, and flag stale reports. Distinguish outbound passengers who may need an overnight stay, canceled inbound flights that may reduce arrivals, and delays that may only shift check-in time. Do not equate flight cancellations or aircraft seats with rooms needed. List verified inventory, existing airline allotments, shuttle limits, unknowns, and scenarios using only [approved assumptions]. Draft options for reservations, staffing, and revenue review, with an owner and next verification time. Do not contact passengers or airlines, change prices, or confirm rooms.",
    "steps": [
      "Choose a relevant airport and a short monitoring window based on the property’s location and guest mix.",
      "Verify the flight status and time zone; reconcile codeshares and distinguish canceled flights from delayed ones.",
      "Compare incoming guest exposure and actual same-day booking pickup with rooms available and existing commitments.",
      "Have reservations and the duty manager review rooms, transport, staffing, and airline arrangements before accepting any new block.",
      "Refresh the brief as flights recover and compare projected stays with actual check-ins and cancellations."
    ],
    "review": "Revenue and operations staff verify current availability, guest impact, airline commitments, and applicable property pricing rules before acting.",
    "limits": [
      "A canceled flight may reduce demand at the destination while creating possible overnight demand elsewhere. Many passengers are local or rebooked the same day.",
      "Flight feeds can lag. A flight alert does not establish hotel demand or guarantee an airline will pay for accommodation.",
      "This is an editorial workflow supported by flight-data capabilities, not a reproduced hotel deployment."
    ],
    "measure": "False alerts, update freshness, actual disruption-related check-ins, missed arrivals, and room-night forecast error.",
    "sources": [
      {
        "title": "AeroAPI flight status and alert delivery",
        "url": "https://www.flightaware.com/commercial/aeroapi/",
        "author": "FlightAware",
        "note": "The product documents flight data and configurable alerts. It does not validate our hotel-demand assumptions or identify travelers needing rooms.",
        "kind": "Vendor example"
      },
      {
        "title": "Event signals for hotel revenue management",
        "url": "https://www.predicthq.com/events/hotel-revenue-management",
        "author": "PredictHQ",
        "note": "The provider identifies airport delays and other events as potential hotel-demand inputs. This guide adds property-specific validation and staff decisions.",
        "kind": "Vendor example"
      }
    ],
    "reviewed": "2026-10-12"
  },
  {
    "id": "airline-capacity-demand",
    "title": "Track new airline routes and seat-capacity changes",
    "description": "Use changes in inbound flights as an early signal for hotel forecasts, source-market campaigns, and staffing plans.",
    "category": "Revenue management",
    "tool": "AI assistant + verified data",
    "toolUrl": null,
    "level": "Intermediate",
    "evidence": "Editorial adaptation",
    "output": "A weekly airport and source-market brief showing capacity changes, hotel booking evidence, and forecast assumptions.",
    "inputs": [
      "Two dated airline schedule snapshots with operating carrier, route, flight date, and seats when available",
      "Property booking pace by stay week and aggregated source market, cancellations, and length of stay",
      "Current hotel forecast, airport catchment, and comparable seasonal periods"
    ],
    "connections": "Use official route announcements for discovery and licensed schedules or capacity data for comparisons. OAG describes hospitality use of schedules and seats data, including a Pestana Hotel Group example. Join these signals to approved PMS or BI summaries.",
    "prompt": "Compare [old snapshot] with [new snapshot] for [destination airport] and [travel weeks]. Separate new routes, frequency changes, aircraft seat changes, and canceled service. Deduplicate operating flights and codeshares. Compute inbound seat changes only where seat data is provided; mark unknowns rather than guessing. Compare the affected weeks and source markets with [hotel booking pace]. Distinguish scheduled capacity from passengers, destination visitors, and hotel room nights. Show scenarios only with [approved load, visitor-share, hotel-capture, party-size, and length-of-stay assumptions], with formulas and uncertainty. Flag connections and outbound resident travel that may weaken the signal. Return forecast-review and campaign questions, not automatic rate or budget changes.",
    "steps": [
      "Select airports and origin markets that plausibly contribute guests to the property.",
      "Compare schedule snapshots for the same future travel weeks; reconcile operating flights and seat definitions.",
      "Join changes to hotel pickup and cancellations by stay week, aligning seasonal and holiday comparisons.",
      "Review scenario assumptions and any marketing or rate response with revenue and sales staff.",
      "Track schedule revisions and test whether the added signal improves forecasts on later, unseen dates."
    ],
    "review": "Revenue staff check operating dates, seat counts, source-market definitions, and scenario conversions; commercial owners approve any campaign or price changes.",
    "limits": [
      "Seats are supply, not passengers or room nights. Added capacity can serve connections, residents, or travelers staying elsewhere.",
      "New services may be delayed, canceled, or substituted; announced capacity can change before departure.",
      "A generic AI prompt is not a calibrated demand model. Compare with a simple existing forecast before relying on it."
    ],
    "measure": "Weekly room-night forecast error versus the prior model, stale schedule alerts, and scenario accuracy by source market.",
    "sources": [
      {
        "title": "Flight data for hospitality, including Pestana Hotel Group",
        "url": "https://www.oag.com/hospitality",
        "author": "OAG",
        "note": "OAG describes using flight schedules and capacity for hospitality decisions and a Pestana integration with Power BI. Our weekly review workflow is an editorial adaptation, not a copy of that implementation.",
        "kind": "Vendor example"
      }
    ],
    "reviewed": "2026-10-12"
  }
]