Implementation Guide: Handle guest messaging end-to-end within defined scope

Hardware Procurement

Deployed at front desk (2), concierge station (1), and housekeeping manager (1) for accessing the AI messaging platform's unified escalation inbox. Staff use these to monitor AI conversations, handle escalated requests, and manage guest check-in workflows.

Software Procurement

Market-leading AI guest messaging platform for mid-market and chain properties. Automates 80%+ of guest communications across SMS, email, webchat, and voice. Includes AI Voice for phone call handling, dynamic upsell engine, contactless check-in, and white-label branding. Best for properties needing voice + text combined or chain-wide standardization.

Prerequisites

• Active cloud-based PMS with API access enabled (Oracle OPERA Cloud, Mews, Cloudbeds, roomMaster, or StayNTouch). If using legacy on-premises OPERA 5, confirm VPN connectivity and API gateway availability before proceeding.
• Minimum 100 Mbps dedicated business internet connection with 99.9%+ uptime SLA. A secondary failover ISP is strongly recommended for properties where messaging uptime is business-critical.
• Existing hotel phone number(s) that can be ported or forwarded for SMS/voice AI. Confirm with current telco that numbers are SMS-capable and can be ported to Twilio.
• Hotel general manager or operations manager designated as project stakeholder with authority to approve: (a) guest-facing message templates and tone of voice, (b) scope of autonomous AI decisions (e.g., max comp value, allowed request types), (c) escalation policies and staff response time SLAs.
• Complete inventory of hotel FAQs, policies, and operational information: check-in/check-out times, cancellation policy, parking details, WiFi instructions, restaurant hours, pool/gym hours, pet policy, smoking policy, local attraction recommendations, shuttle schedules, room service menu, and any seasonal/event-specific information.
• PMS administrator credentials or API key provisioning access. The MSP technician will need: (a) API client ID and secret for the PMS, (b) webhook endpoint configuration capability, (c) read access to reservations, guest profiles, room status, and folios.
• Credit card or billing account for Twilio (or confirmation that the AI messaging platform includes messaging infrastructure in their pricing).
• For WhatsApp channel: Facebook Business Manager account linked to the hotel, with a verified business phone number. WhatsApp Business API approval typically takes 1–5 business days.
• Staff email addresses for all users who will access the AI messaging escalation inbox (typically 4–8 users: front desk agents, concierge, housekeeping supervisor, GM, duty manager).
• Physical site survey completed: Ethernet drops at front desk, concierge, and network closet confirmed. Power outlets available for AP placement locations. Network closet has adequate ventilation and power.
• PCI DSS 4.0.1 self-assessment questionnaire (SAQ) completed or scheduled. Confirm current PCI compliance status and identify any gaps that this deployment must not worsen.
• Written guest data privacy policy reviewed by hotel management or legal counsel, covering: AI chatbot disclosure, data collection scope, retention periods, and opt-out mechanisms. Template will be provided during implementation.

Installation Steps

Step 1: Network Infrastructure Assessment and VLAN Design

Conduct a physical and logical network assessment of the hotel property. Document existing network topology, internet service details, switch infrastructure, and wireless coverage. Design the VLAN architecture that will segment traffic for PCI DSS compliance and optimal performance. Create four VLANs: Guest WiFi (VLAN 100), Staff Operations (VLAN 200), PMS/Payment Systems (VLAN 300), and IoT/Smart Locks (VLAN 400).

Network discovery scan, bandwidth baseline, and VLAN design reference

Step 2: Install and Configure FortiGate 40F Firewall

Rack-mount or desktop-deploy the FortiGate 40F firewall. Connect WAN port to the ISP modem/handoff. Configure LAN interfaces with VLAN sub-interfaces. Set up firewall policies that allow the AI messaging platform's cloud endpoints (HTTPS/443 outbound), PMS API traffic, and block unauthorized inter-VLAN routing. Enable IPS and web filtering. Register the device with FortiCloud for MSP remote management.

# Initial setup via console cable (115200 baud) or default IP https://192.168.1.99
# Set hostname and management credentials
config system global
  set hostname Hotel-FW-01
  set timezone 12
  set admin-sport 8443
end

# Configure WAN interface (adjust for ISP settings)
config system interface
  edit port1
    set alias WAN
    set mode dhcp
    set allowaccess ping https ssh
  next
end

# Configure LAN trunk interface with VLANs
config system interface
  edit port2
    set alias LAN-Trunk
    set type physical
  next
  edit VLAN100-Guest
    set vdom root
    set ip 10.100.0.1 255.255.0.0
    set allowaccess ping
    set interface port2
    set vlanid 100
  next
  edit VLAN200-Staff
    set vdom root
    set ip 10.200.0.1 255.255.255.0
    set allowaccess ping https ssh
    set interface port2
    set vlanid 200
  next
  edit VLAN300-PMS
    set vdom root
    set ip 10.300.0.1 255.255.255.0
    set allowaccess ping
    set interface port2
    set vlanid 300
  next
  edit VLAN400-IoT
    set vdom root
    set ip 10.400.0.1 255.255.255.0
    set allowaccess ping
    set interface port2
    set vlanid 400
  next
end

# DHCP server for Staff VLAN
config system dhcp server
  edit 1
    set interface VLAN200-Staff
    config ip-range
      edit 1
        set start-ip 10.200.0.100
        set end-ip 10.200.0.200
      next
    end
    set default-gateway 10.200.0.1
    set dns-server1 8.8.8.8
    set dns-server2 8.8.4.4
  next
end

# Firewall policy: Staff VLAN -> Internet (for AI messaging platform access)
config firewall policy
  edit 1
    set name Staff-to-Internet
    set srcintf VLAN200-Staff
    set dstintf port1
    set srcaddr all
    set dstaddr all
    set action accept
    set schedule always
    set service HTTPS HTTP DNS
    set nat enable
    set utm-status enable
    set ips-sensor default
    set ssl-ssh-profile certificate-inspection
  next
end

# Firewall policy: PMS VLAN -> Internet (restricted to PMS cloud + AI platform APIs)
config firewall policy
  edit 2
    set name PMS-to-Cloud
    set srcintf VLAN300-PMS
    set dstintf port1
    set srcaddr all
    set dstaddr all
    set action accept
    set schedule always
    set service HTTPS
    set nat enable
    set utm-status enable
    set ips-sensor default
  next
end

# DENY Guest VLAN -> Staff/PMS VLANs (critical for PCI)
config firewall policy
  edit 10
    set name Block-Guest-to-Internal
    set srcintf VLAN100-Guest
    set dstintf VLAN200-Staff VLAN300-PMS VLAN400-IoT
    set srcaddr all
    set dstaddr all
    set action deny
    set schedule always
    set service ALL
  next
end

# Register with FortiCloud
exec forticloud register

Step 3: Deploy and Configure Ubiquiti Network Switch and Access Points

Install the USW-Pro-24-PoE switch in the network closet. Connect uplink to the FortiGate LAN trunk port. Configure VLANs on switch ports matching the firewall VLAN design. Deploy U6-Pro access points throughout the property, connecting each to PoE switch ports. Adopt all devices into the UniFi Network controller and configure wireless networks per VLAN.

• Settings > Networks > Create New Network: Name: Guest-WiFi | VLAN ID: 100 | Subnet: 10.100.0.0/16 | DHCP: FortiGate (10.100.0.1)
• Settings > Networks > Create New Network: Name: Staff-Ops | VLAN ID: 200 | Subnet: 10.200.0.0/24 | DHCP: FortiGate (10.200.0.1)
• Settings > Networks > Create New Network: Name: PMS-Net | VLAN ID: 300 | Subnet: 10.300.0.0/24 | DHCP: FortiGate (10.300.0.1)
• Settings > Networks > Create New Network: Name: IoT-Net | VLAN ID: 400 | Subnet: 10.400.0.0/24 | DHCP: FortiGate (10.400.0.1)
• Settings > WiFi > Create New WiFi Network: SSID: [HotelName]-Guest | Network: Guest-WiFi | Security: WPA2 | Guest Policy: ON
• Settings > WiFi > Create New WiFi Network: SSID: [HotelName]-Staff | Network: Staff-Ops | Security: WPA2-Enterprise or WPA2-PSK

Switch Port Profiles:
  Ports 1-2:   Trunk (All VLANs)            - uplinks
  Ports 3-10:  Staff-Ops (VLAN 200 native)  - front desk PCs, tablets
  Ports 11-16: PMS-Net (VLAN 300 native)    - PMS terminals, payment devices
  Ports 17-24: Trunk (VLAN 100, 200)        - AP connections (tagged Guest + Staff WiFi)

Step 4: Provision Twilio Account and Phone Number

Create a Twilio account (or use existing MSP sub-account structure), provision a local or toll-free phone number for the hotel, configure the number for SMS capability, and set up the messaging service that the AI platform will use. If the hotel wants to keep their existing number, initiate a number port to Twilio.

Account SID: ACXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
Auth Token: (from sub-account)
Phone Number: +1XXXXXXXXXX
Messaging Service SID: MGXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

Step 5: Set Up PMS API Access and Test Connectivity

Work with the hotel's PMS administrator to enable API access for the AI messaging platform. Obtain API credentials, configure webhook endpoints, and test basic data retrieval (reservations, guest profiles, room status). This is the most critical integration and the most likely source of delays.

For Mews PMS

curl -X POST https://api.mews.com/api/connector/v1/reservations/getAll \
  -H 'Content-Type: application/json' \
  -d '{
    "ClientToken": "YOUR_CLIENT_TOKEN",
    "AccessToken": "YOUR_ACCESS_TOKEN",
    "Client": "AI-Messaging-Integration",
    "StartUtc": "2025-01-01T00:00:00Z",
    "EndUtc": "2025-01-02T00:00:00Z"
  }'

For Oracle OPERA Cloud

curl -X POST https://[region].oraclehospitality.com/oauth/v1/tokens \
  -H 'Content-Type: application/x-www-form-urlencoded' \
  -d 'grant_type=client_credentials&client_id=YOUR_CLIENT_ID&client_secret=YOUR_CLIENT_SECRET'

For Cloudbeds

curl -X GET 'https://api.cloudbeds.com/api/v1.2/getReservations?status=confirmed' \
  -H 'Authorization: Bearer YOUR_API_KEY'

Verify You Can Retrieve

• Current reservations (today + next 7 days)
• Guest profile data (name, email, phone)
• Room status (clean, dirty, occupied, vacant)
• Folio/billing data (for upsell context)

Step 6: Deploy and Configure AI Guest Messaging Platform (Akia - Primary)

Create the hotel's Akia account, configure the property profile, connect the PMS integration, set up messaging channels (SMS, webchat, WhatsApp), customize the AI knowledge base, and configure automated message flows for the guest journey. This is the core of the deployment and will take the most configuration time.

<!-- install before </body> on hotel website -->

<script src='https://chat.akia.com/widget/[PROPERTY_ID].js'></script>

Step 7: Configure Staff Escalation Inbox and User Accounts

Create user accounts for all hotel staff who will handle escalated conversations. Configure the unified inbox interface, set up notification preferences, define shift schedules for routing, and install the Akia app on staff tablets.

Step 8: Install Webchat Widget on Hotel Website

Deploy the AI messaging webchat widget on the hotel's website. This provides a direct channel for potential guests browsing the website to interact with the AI agent for pre-booking questions, availability inquiries, and direct booking assistance.

<script>
  (function(w,d,s,o,f,js,fjs){
    w['AkiaWidget']=o;w[o]=w[o]||function(){
    (w[o].q=w[o].q||[]).push(arguments)};
    js=d.createElement(s),fjs=d.getElementsByTagName(s)[0];
    js.id=o;js.src=f;js.async=1;fjs.parentNode.insertBefore(js,fjs);
  }(window,document,'script','akia','https://chat.akia.com/widget/PROPERTY_ID.js'));
  akia('init', { propertyId: 'PROPERTY_ID', theme: 'custom', primaryColor: '#1a5276' });
</script>

echo '<script src="https://chat.akia.com/widget/PROPERTY_ID.js"></script>' >> footer-include.html

Step 9: Build and Test AI Knowledge Base

This is the most labor-intensive and highest-impact step. Populate the AI agent's knowledge base with comprehensive hotel-specific information so it can accurately and helpfully respond to guest inquiries autonomously. Work directly with hotel operations leadership to capture all necessary information.

• In Akia Dashboard > AI Settings > Knowledge Base, create the following categories and articles:
• CATEGORY: Property Information — Articles: Hotel Overview (address, phone, description, star rating, brand); Check-in / Check-out Times (standard + early/late policies + fees); Parking (valet/self-park, rates, EV charging, height restrictions); WiFi (network name, password, troubleshooting steps); Pet Policy (allowed?, fee, weight limit, designated areas); Smoking Policy (smoke-free property? designated areas?); ADA Accessibility (accessible rooms, facilities, services)
• CATEGORY: Rooms & Amenities — Articles: Room Types (descriptions, bed configurations, max occupancy, views); In-Room Amenities (coffee maker, mini bar, safe, iron, hair dryer); Room Requests (extra pillows, towels, rollaway bed, crib - and how to request); Pool & Hot Tub (hours, rules, towel policy, seasonal availability); Fitness Center (hours, equipment list, any personal trainer services); Spa (services menu, hours, booking process, pricing); Business Center (hours, printing, fax, meeting rooms); Laundry (self-service location, valet laundry pricing and turnaround)
• CATEGORY: Dining — Articles: On-Site Restaurant (name, cuisine, hours, dress code, reservation policy); Room Service (hours, menu highlights, ordering instructions); Bar/Lounge (hours, happy hour details, dress code); Breakfast (included?, hours, location, menu type); Coffee Shop (hours, location); Nearby Restaurants (top 5–10 with cuisine type, distance, price range)
• CATEGORY: Local Area & Activities — Articles: Getting Here (directions from airport, train station, highway); Airport Shuttle (schedule, booking, cost, pickup location); Local Attractions (top 10 with descriptions, distance, hours, ticket prices); Shopping (nearby malls, boutiques, convenience stores); Nightlife (bars, clubs, entertainment venues); Events & Conferences (nearby convention center, current events); Weather & What to Pack (seasonal guidance)
• CATEGORY: Services & Requests — Articles: Housekeeping Requests (extra cleaning, towels, DND preferences); Maintenance Issues (how to report, what constitutes emergency); Lost & Found (process, contact, timeline); Concierge Services (tour booking, restaurant reservations, tickets); Transportation (taxi, Uber/Lyft, car rental, shuttle schedules); Package & Mail Handling
• CATEGORY: Policies — Articles: Cancellation Policy (by rate type: flexible, non-refundable, group); Payment Methods Accepted; Incidental Hold / Authorization Amount; Noise Policy & Quiet Hours; Visitor/Guest Policy; Maximum Occupancy Rules
• CATEGORY: Troubleshooting — Articles: WiFi Not Working (step-by-step troubleshooting); Room Key Not Working (quick fixes + front desk contact); AC/Heat Not Working (thermostat instructions + maintenance request); TV/Entertainment System (channel guide, streaming instructions); Safe Instructions (how to set/reset code)

• After populating, test with at least 50 sample questions:
• 'What time is checkout?' → expects: check-out time + late checkout offer
• 'Do you have parking?' → expects: parking options + rates
• 'Can I bring my dog?' → expects: pet policy + fee details
• 'What restaurants are nearby?' → expects: local dining recommendations
• 'I need more towels' → expects: creates housekeeping request + ETA

Step 10: Configure Automated Upsell and Revenue Workflows

Set up automated upsell offers that the AI agent will present to guests at optimal moments during their journey. These offers generate incremental revenue for the hotel and demonstrate clear ROI for the AI messaging investment.

Step 11: Staff Training and Soft Launch

Conduct training sessions with hotel staff on the AI messaging system. Run a soft launch with a limited room block to validate the system in production before full rollout. Monitor AI accuracy, staff response times, and guest satisfaction during the soft launch period.

Training Session Agenda (2 hours)

# Day 1-2: AI messaging enabled for 20% of arriving guests (specific room block)
# Day 3-4: Expand to 50% of arrivals
# Day 5-7: Expand to 100% of arrivals

Soft Launch Monitoring Metrics & Targets

Step 12: Full Deployment, Analytics Configuration, and Go-Live

After successful soft launch validation, enable AI messaging for all guests across all channels. Configure analytics dashboards and automated reporting for the hotel management team and MSP monthly review. Establish baseline KPIs and success criteria.

Custom AI Components

Hotel Knowledge Base Schema

Type: prompt

Structured knowledge base template that defines how hotel-specific information is organized and fed to the AI agent. This ensures comprehensive coverage of all guest inquiry categories and enables consistent AI responses. Used to populate the AI messaging platform's knowledge base in Step 9.

Implementation:

# Hotel Knowledge Base Schema for AI Guest Messaging Agent
# Use this template to gather all property-specific information
# Each section maps to a knowledge base category in the AI platform

---

SECTION 1: PROPERTY OVERVIEW

• Property Name: [Full legal name and brand affiliation]
• Address: [Full address with zip code]
• Phone: [Main number]
• Email: [General email]
• Star Rating: [Official rating]
• Brief Description: [2-3 sentence elevator pitch for the property]
• Unique Selling Points: [What makes this property special - list 3-5]
• Total Rooms: [Number]
• Number of Floors: [Number]
• Year Built / Last Renovated: [Year]

SECTION 2: CHECK-IN / CHECK-OUT

• Standard Check-in Time: [e.g., 3:00 PM]
• Standard Checkout Time: [e.g., 11:00 AM]
• Early Check-in Policy: [Available? Fee? How early? Must request in advance?]
• Late Checkout Policy: [Available? Fee? How late? Subject to availability?]
• Digital/Mobile Check-in: [Available? Process description]
• ID Requirements: [Government-issued photo ID, credit card at check-in]
• Age Requirement: [Minimum age to check in, typically 18 or 21]
• After-Hours Check-in: [Process if arriving after front desk hours]

SECTION 3: ROOM TYPES (repeat for each room type)

• Room Type Name: [e.g., Deluxe King]
• Description: [2-3 sentences]
• Bed Configuration: [e.g., 1 King bed]
• Max Occupancy: [Adults + Children]
• Room Size: [Square feet/meters]
• View: [City/Ocean/Garden/Courtyard/None]
• Floor(s): [Which floors]
• Key Amenities: [List all in-room amenities]
• ADA Accessible Version: [Available? How many?]

SECTION 4: IN-ROOM AMENITIES (master list)

SECTION 5: WIFI INSTRUCTIONS

• Network Name: [SSID]
• Password: [If applicable]
• Connection Steps: [Step-by-step for guests]

Troubleshooting: [Common issues and fixes]

• Bandwidth: [Adequate for streaming? Any restrictions?]
• Premium WiFi: [Available? Cost?]

SECTION 6: DINING

On-Site Restaurant(s) (repeat for each)

• Name: [Restaurant name]
• Cuisine Type: [e.g., Mediterranean]
• Hours: [Breakfast/Lunch/Dinner with times]
• Dress Code: [Smart casual, etc.]
• Reservation Required: [Yes/No, how to book]
• Menu Highlights: [3-5 popular items]
• Dietary Accommodations: [Vegetarian, vegan, gluten-free, kosher, halal]

Room Service

• Hours: [24-hour or limited]
• Ordering: [Dial ext, use app, text us]
• Delivery Time: [Typical wait]
• Tray Pickup: [Instructions for when done]

Breakfast

• Included: [Yes/No, which rate plans]
• Type: [Buffet/Continental/A la carte]
• Hours: [Time range]
• Location: [Restaurant name/floor]

Bar/Lounge

• Name: [Bar name]
• Hours: [Opening/closing]
• Happy Hour: [Times, specials]
• Dress Code: [If applicable]

SECTION 7: HOTEL AMENITIES

Pool

• Type: [Indoor/Outdoor/Both]
• Hours: [Opening/closing]
• Seasonal: [Open year-round or seasonal dates]
• Towels: [Provided at pool or bring from room]
• Rules: [No glass, lifeguard on duty?, children policy]
• Hot Tub: [Available? Hours? Capacity?]

Fitness Center

• Hours: [e.g., 24 hours with room key access]
• Equipment: [List major equipment]
• Towels/Water: [Provided?]
• Personal Trainer: [Available? How to book?]

Spa

• Hours: [Operating hours]
• Services: [Top services with approximate pricing]
• Booking: [How to book - phone, app, concierge]
• Cancellation: [Spa cancellation policy]

Business Center

• Hours: [Operating hours]
• Services: [Printing, fax, computer access]
• Meeting Rooms: [Available? Capacity? Booking process?]

SECTION 8: PARKING & TRANSPORTATION

• Self-Parking: [Available? Rate per day? Location?]
• Valet Parking: [Available? Rate per day? Hours?]
• EV Charging: [Available? Number of stations? Cost? Type (Level 2/DC Fast)?]
• Height Restrictions: [Garage clearance if applicable]
• Oversized Vehicle: [Policy for RVs, buses]
• Airport Shuttle: [Available? Schedule? Cost? Booking required?]
• Local Shuttle: [Complimentary area shuttle? Radius? Hours?]
• Taxi/Rideshare: [Pickup location, estimated costs to key destinations]
• Car Rental: [On-site or nearby? Partner company?]
• Public Transit: [Nearest stop/station, directions]

SECTION 9: LOCAL AREA

Nearby Attractions (list top 10)

• Name: [Attraction]
• Type: [Museum/Park/Entertainment/Shopping]
• Distance: [Miles/minutes from hotel]
• Hours: [Operating hours]
• Ticket Price: [Approximate, or if hotel offers discounted tickets]
• Getting There: [Walk/drive/shuttle instructions]

Nearby Restaurants (top 10 off-property)

• Name: [Restaurant]
• Cuisine: [Type]
• Price Range: [$/$$/$$$]
• Distance: [Walking/driving]
• Reservation Recommended: [Yes/No]
• Hotel Guest Discount: [If applicable]

Nearby Essentials

• Pharmacy: [Nearest, with address and distance]
• Grocery Store: [Nearest]
• Convenience Store: [Nearest, or hotel gift shop]
• Hospital/Urgent Care: [Nearest, with address and phone]

SECTION 10: POLICIES

• Cancellation Policy: [By rate type] - Flexible: [e.g., Free cancellation until 6 PM day of arrival] - Non-Refundable: [No cancellation/modification] - Group: [Separate policy]
• Payment: [Accepted cards, cash policy]
• Incidental Hold: [Amount, when charged, when released]
• Pet Policy: [Allowed? Fee? Weight limit? Breed restrictions? Pet-free floors?]
• Smoking: [Property-wide non-smoking? Designated areas? Fee for violations?]
• Noise/Quiet Hours: [e.g., 10 PM - 8 AM]
• Visitor Policy: [Non-guests allowed in rooms? Lobby access?]
• Damage Policy: [How assessed and charged]

SECTION 11: SPECIAL REQUESTS & HOUSEKEEPING

• Extra Towels: [How to request, typical delivery time]
• Extra Pillows/Blankets: [Available types, request process]
• Rollaway Bed: [Available? Fee? Must request in advance?]
• Crib: [Available? Fee? Request process]
• Connecting Rooms: [Available? How to request]
• Hypoallergenic Bedding: [Available?]
• Housekeeping Schedule: [Daily? Time? Opt-out option?]
• Do Not Disturb: [How to set, policy for extended DND]
• Turndown Service: [Available? Request process?]

SECTION 12: SAFETY & EMERGENCY

• Emergency Number: [Dial 911 or front desk extension]
• Fire Exit: [General directions, noted on room door]
• First Aid: [Available at front desk]
• AED Location: [If applicable]
• Security: [24-hour? Contact method?]
• Safe Location: [In-room safe instructions]

Pre-Arrival Welcome Workflow

Type: workflow

Automated workflow that triggers 48 hours before guest arrival, sends a personalized welcome message via SMS, offers digital check-in, and presents early check-in and room upgrade upsell opportunities based on PMS reservation data and real-time room availability.

Implementation:

# Pre-Arrival Welcome Workflow
# Platform: Akia (or equivalent AI messaging platform)
# Trigger: PMS reservation sync detects arrival date = current_date + 2 days

Workflow Definition (YAML-style configuration)

# full YAML-style configuration including trigger, steps, error handling,
# and analytics

workflow:
  name: pre_arrival_welcome
  trigger:
    type: time_based
    condition: reservation.arrival_date == today() + 2_days
    time: 10:00_AM_property_timezone
    filter:
      - reservation.status == 'confirmed'
      - reservation.channel != 'walk_in'
      - guest.phone_number IS NOT NULL
      - guest.opted_out_messaging != true

  steps:
    - step: 1
      action: send_message
      channel: sms  # primary; fallback to email if no phone
      template: |
        Hi {{guest.first_name}}! 🏨 We're looking forward to welcoming you
        to {{property.name}} on {{reservation.arrival_date_formatted}}!

        Your reservation details:
        📅 {{reservation.arrival_date_formatted}} - {{reservation.departure_date_formatted}}
        🛏️ {{reservation.room_type_name}}
        👥 {{reservation.adult_count}} adult(s){{#if reservation.child_count}}, {{reservation.child_count}} child(ren){{/if}}

        ✅ Save time at arrival — complete your digital check-in now:
        {{digital_checkin_link}}

        Reply to this message anytime with questions about your stay.
        We're here 24/7! 💬
      delay: 0

    - step: 2
      action: check_upsell_eligibility
      delay: 30_minutes_after_step_1
      conditions:
        early_checkin:
          check: room_availability(reservation.room_type, reservation.arrival_date, before=reservation.standard_checkin_time)
          if_true: go_to_step_3a
        room_upgrade:
          check: upgrade_availability(reservation.room_type, reservation.arrival_date, reservation.departure_date)
          if_true: go_to_step_3b
        both_available:
          if_true: go_to_step_3c  # combined offer
        neither_available:
          if_true: end_workflow

    - step: 3a
      action: send_upsell_offer
      type: early_checkin
      template: |
        By the way, {{guest.first_name}} — would you like to check in
        early on {{reservation.arrival_date_formatted}}?

        ⏰ Check in as early as {{early_checkin_time}} for just ${{early_checkin_price}}.

        Reply YES to add early check-in, or NO THANKS.
      on_response:
        'YES':
          - create_upsell_charge(early_checkin_price)
          - send_payment_link()
          - update_reservation(early_checkin=true)
          - send_confirmation: "Early check-in confirmed! See you at {{early_checkin_time}}. 🎉"
        'NO THANKS':
          - send_acknowledgment: "No problem! Standard check-in is at {{standard_checkin_time}}."
        no_response_24h:
          - end_workflow

    - step: 3b
      action: send_upsell_offer
      type: room_upgrade
      template: |
        Great news, {{guest.first_name}}! 🌟 We have a {{upgrade_room_type}}
        available for your dates.

        Upgrade and enjoy:
        {{upgrade_benefits_list}}

        Just ${{upgrade_price_per_night}}/night extra ({{upgrade_total_price}} total).

        Reply UPGRADE to secure it, or let me know if you have questions!
      on_response:
        'UPGRADE':
          - create_upsell_charge(upgrade_total_price)
          - send_payment_link()
          - update_reservation(room_type=upgrade_room_type)
          - send_confirmation: "Awesome! You've been upgraded to a {{upgrade_room_type}}! 🎊"
        decline:
          - send_acknowledgment: "No worries! Your {{original_room_type}} is a great choice."

    - step: 3c
      action: send_combined_offer
      template: |
        {{guest.first_name}}, we have two options to enhance your stay:

        1️⃣ Early Check-in at {{early_checkin_time}} — ${{early_checkin_price}}
        2️⃣ Room Upgrade to {{upgrade_room_type}} — ${{upgrade_price_per_night}}/night
        3️⃣ Both! — ${{combined_price}} (save ${{combined_savings}}!)

        Reply 1, 2, 3, or PASS.
      on_response:
        '1': trigger_step_3a_acceptance
        '2': trigger_step_3b_acceptance
        '3': trigger_combined_acceptance
        'PASS': send_acknowledgment

  error_handling:
    sms_delivery_failure:
      - retry_once_after: 2_hours
      - if_still_failed: send_via_email
    pms_api_error:
      - log_error
      - alert_msp_team
      - skip_upsell_steps
    payment_failure:
      - notify_guest: "No worries — we can add this at check-in instead."
      - flag_for_front_desk

  analytics:
    track:
      - message_delivered: boolean
      - message_read: boolean (if available)
      - digital_checkin_completed: boolean
      - upsell_offered: type
      - upsell_accepted: boolean
      - upsell_revenue: amount
      - response_time: seconds

In-Stay Request Router Agent

Type: agent

Autonomous AI agent that receives guest messages during their stay, classifies the intent, and either handles the request directly (informational queries), creates operational tasks (housekeeping, maintenance), triggers upsell flows (spa, dining), or escalates to human staff (complaints, billing, safety). This is the core autonomous agent that handles 80%+ of in-stay communications.

Implementation:

Agent System Prompt (configure in platform's AI settings)

Intent Classification & Routing Rules

intent_routing:
  
  # TIER 1: AI handles autonomously (no human needed)
  informational:
    intents:
      - hotel_hours        # "What time does the pool close?"
      - amenity_info       # "Do you have a gym?"
      - wifi_help          # "What's the WiFi password?"
      - directions         # "How do I get to the airport?"
      - local_recommendations  # "Any good Italian restaurants nearby?"
      - policy_question    # "What's your cancellation policy?"
      - room_info          # "Does my room have a coffee maker?"
      - checkout_info      # "What time is checkout?"
      - parking_info       # "Where do I park?"
      - dining_info        # "What's on the breakfast menu?"
    action: respond_from_knowledge_base
    fallback: escalate_if_confidence < 0.7
  
  # TIER 2: AI creates task + confirms with guest
  service_request:
    intents:
      - extra_towels       # "Can I get more towels?"
      - extra_pillows      # "I need another pillow"
      - extra_amenities    # "Can I get more shampoo?"
      - room_cleaning      # "Can someone clean my room?"
      - turndown_service   # "Can I get turndown service?"
    action: create_housekeeping_task
    response_template: |
      Of course, {{guest.first_name}}! I've sent your request for
      {{request_item}} to our housekeeping team. They'll be at room
      {{reservation.room_number}} within {{estimated_time}}. 🙂
    estimated_times:
      extra_towels: "15-20 minutes"
      extra_pillows: "15-20 minutes"
      extra_amenities: "15-20 minutes"
      room_cleaning: "30-45 minutes"
      turndown_service: "by 8:00 PM"
  
  maintenance_request:
    intents:
      - ac_not_working     # "My AC isn't working"
      - tv_not_working     # "TV won't turn on"
      - plumbing_issue     # "Toilet is running"
      - light_burned_out   # "Bathroom light is out"
      - key_not_working    # "My room key doesn't work"
      - noise_complaint    # "The room next door is loud"
    action: create_maintenance_task
    pre_action: attempt_troubleshooting_first
    troubleshooting:
      ac_not_working: |
        I'm sorry about that! Let's try a quick fix first:
        1. Check if the thermostat is set to COOL mode
        2. Try setting the temperature 5 degrees lower
        3. Make sure the unit isn't blocked by curtains
        
        Did that help, or should I send our maintenance team?
      tv_not_working: |
        Let me help troubleshoot:
        1. Press the power button on the remote (not the TV)
        2. If no response, try the TV's power button on the bottom edge
        3. Check that the TV is plugged in behind the console
        
        Still not working? I'll send someone right up!
      key_not_working: |
        Sorry about that! Sometimes room keys demagnetize near phones.
        Please visit the front desk and they'll make you a new key
        right away. They're available 24/7! 🔑
    post_troubleshoot_fail:
      action: create_maintenance_task + notify_front_desk
      response: |
        I've notified our maintenance team. Someone will be at room
        {{reservation.room_number}} within {{estimated_time}}. We
        apologize for the inconvenience!
  
  # TIER 3: AI presents upsell + handles transaction
  upsell_opportunity:
    intents:
      - spa_inquiry        # "Do you have a spa?"
      - dining_reservation # "Can I book dinner?"
      - late_checkout_ask  # "Can I check out late?"
      - upgrade_inquiry    # "Can I upgrade my room?"
      - activity_inquiry   # "What is there to do around here?"
    action: present_relevant_upsell + provide_info
    context_rules:
      - If guest asks about spa -> provide spa info + offer booking link
      - If guest asks about late checkout -> check availability + offer
      - If guest mentions special occasion -> suggest celebration package
  
  # TIER 4: MUST escalate to human
  escalation_required:
    intents:
      - billing_dispute    # "My bill is wrong"
      - rate_complaint     # "I was quoted a different rate"
      - compensation_request  # "I want a refund"
      - safety_concern     # "I feel unsafe" / "There's a fire"
      - medical_emergency  # "I need a doctor" / "Someone is hurt"
      - harassment_report  # "Another guest is harassing me"
      - severe_complaint   # "This is unacceptable" + negative sentiment
      - human_request      # "I want to talk to a person"
      - legal_threat       # "I'm going to sue" / "lawyer"
    action: immediate_escalation
    response: |
      I understand, {{guest.first_name}}. Let me connect you with
      our {{department}} team right away. Someone will be with you
      within {{sla_time}}.
    priority_levels:
      safety_emergency: P1  # Immediate + phone call to duty manager
      billing_complaint: P2  # Within 3 minutes
      general_escalation: P3  # Within 5 minutes
      human_preference: P3   # Within 5 minutes

Sentiment Detection Thresholds

sentiment_monitoring:
  positive_threshold: 0.6   # Log as positive interaction
  neutral_range: 0.3-0.6    # Continue normally
  negative_threshold: 0.3   # Flag for review
  critical_threshold: 0.15  # Auto-escalate regardless of intent
  consecutive_negative: 2   # If 2+ negative messages in a row, escalate

Conversation Guardrails

guardrails:
  max_ai_turns_without_resolution: 5  # Escalate if AI can't resolve in 5 turns
  max_conversation_duration: 30_minutes  # Alert supervisor if open > 30 min
  profanity_handling: acknowledge_frustration_and_escalate
  off_topic_handling: politely_redirect_or_escalate
  pii_detection: never_repeat_credit_card_numbers_in_messages
  language_detection: if_non_english_attempt_response_else_escalate

Post-Stay Review Solicitation Workflow

Type: workflow

Automated workflow that triggers 24 hours after checkout, thanks the guest, solicits feedback through a satisfaction rating, and routes satisfied guests to public review platforms (Google, TripAdvisor) while routing dissatisfied guests to an internal recovery process to prevent negative public reviews.

Implementation:

# Post-Stay Review Solicitation Workflow
# Trigger: 24 hours after PMS checkout event

workflow:
  name: post_stay_review
  trigger:
    type: event_based
    event: reservation.status_changed_to_checked_out
    delay: 24_hours
    filter:
      - reservation.stay_duration >= 1_night
      - guest.phone_number IS NOT NULL
      - guest.opted_out_messaging != true
      - conversation.negative_escalation_unresolved != true
        # Don't solicit review if there's an open complaint

  steps:
    - step: 1
      action: send_thank_you_and_rate
      channel: sms
      template: |
        Hi {{guest.first_name}}, thank you for staying at
        {{property.name}}! 🙏
        
        How was your experience? Reply with a number:
        5 ⭐ - Excellent
        4 ⭐ - Very Good
        3 ⭐ - Good
        2 ⭐ - Fair
        1 ⭐ - Poor
      timeout: 48_hours
      on_no_response: end_workflow

    - step: 2a
      condition: response in [4, 5]
      action: route_to_public_review
      template: |
        Wonderful, thank you! We're so glad you enjoyed your stay! 😊
        
        Would you mind sharing your experience? It really helps
        other travelers:
        
        📝 Google: {{property.google_review_link}}
        📝 TripAdvisor: {{property.tripadvisor_link}}
        
        Thank you, {{guest.first_name}}! We hope to welcome you
        back soon. 🏨
      analytics:
        track: promoter_review_requested

    - step: 2b
      condition: response == 3
      action: acknowledge_and_gather_feedback
      template: |
        Thank you for your feedback, {{guest.first_name}}.
        We always aim for excellence — is there anything
        specific we could have done better?
        
        Your input helps us improve! 🙂
      on_response:
        action: log_feedback
        notify: hotel_gm_email
        respond: |
          Thank you for sharing that. I've passed your feedback
          to our management team. We hope to exceed your
          expectations next time!

    - step: 2c
      condition: response in [1, 2]
      action: service_recovery_escalation
      template: |
        We're truly sorry your experience didn't meet expectations,
        {{guest.first_name}}. Your feedback is important to us.
        
        Could you share what went wrong? Our General Manager
        would like to personally address your concerns.
      on_response:
        action: create_service_recovery_ticket
        assign_to: general_manager
        priority: HIGH
        include: full_conversation_history
        respond: |
          Thank you for letting us know. Our General Manager,
          {{gm.first_name}}, will be reaching out to you within
          24 hours. We take this seriously and want to make it right.
      analytics:
        track: detractor_recovery_initiated

  analytics:
    track:
      - survey_sent: count
      - survey_responded: count_and_rate
      - rating_distribution: [1,2,3,4,5]
      - review_link_clicked: boolean
      - recovery_tickets_created: count
      - nps_equivalent: (5_count + 4_count - 1_count - 2_count) / total

PMS-to-Housekeeping Task Bridge

Type: integration

Integration component that bridges the AI messaging platform with the hotel's housekeeping/task management system. When a guest requests housekeeping services via the AI agent, this integration creates a structured task in the housekeeping system (Quore, Flexkeeping, or native PMS task module), tracks completion status, and sends a confirmation message to the guest when the task is completed.

Implementation:

• PMS-to-Housekeeping Task Bridge
• Implemented via n8n workflow automation or platform-native webhooks

n8n Workflow Configuration

• Trigger: Webhook from AI messaging platform
• URL: https://your-n8n-instance.com/webhook/housekeeping-task
• Method: POST
• Authentication: Header-based API key

Webhook Payload Schema (from AI messaging platform)

{
  "event_type": "task_created",
  "source": "ai_messaging_agent",
  "timestamp": "2025-07-15T14:30:00Z",
  "guest": {
    "id": "guest_12345",
    "first_name": "Sarah",
    "last_name": "Johnson",
    "room_number": "412",
    "reservation_id": "res_67890",
    "phone": "+15551234567",
    "vip_level": "gold"
  },
  "task": {
    "type": "housekeeping",  # housekeeping | maintenance | concierge
    "category": "extra_towels",  # extra_towels | extra_pillows | room_cleaning | amenity_restock | turndown
    "description": "Guest requested extra bath towels",
    "quantity": 2,
    "priority": "normal",  # urgent | high | normal | low
    "requested_time": null,  # null = ASAP, or ISO timestamp
    "original_message": "Can I get a couple more towels please?"
  },
  "conversation_id": "conv_abc123",
  "callback_url": "https://api.akia.com/webhooks/task-update"
}

n8n Workflow Steps

• Node 1: Webhook Trigger
• Type: Webhook
• HTTP Method: POST
• Path: /housekeeping-task
• Authentication: Header Auth (X-API-Key)

• Node 2: Validate Payload
• Type: IF
• Condition: {{$json.task.type}} is not empty AND {{$json.guest.room_number}} is not empty
• True: Continue
• False: Respond with 400 error

• Node 3: Route by Housekeeping System
• Type: Switch
• Based on environment variable: HOUSEKEEPING_SYSTEM

Route A: Quore Integration

• Node 3A: HTTP Request to Quore API
• Method: POST
• URL: https://api.quore.com/v1/tasks
• Headers:
• Authorization: Bearer {{$env.QUORE_API_KEY}}
• Content-Type: application/json

Body:

{
  "property_id": "{{$env.QUORE_PROPERTY_ID}}",
  "task_type": "guest_request",
  "category": "{{$json.task.category}}",
  "room": "{{$json.guest.room_number}}",
  "description": "{{$json.task.description}} (Qty: {{$json.task.quantity}})",
  "priority": "{{$json.task.priority}}",
  "guest_name": "{{$json.guest.first_name}} {{$json.guest.last_name}}",
  "source": "AI Guest Messaging",
  "callback_url": "https://your-n8n-instance.com/webhook/task-completed"
}

Route B: Native PMS Task Module (Mews)

• Node 3B: HTTP Request to Mews API
• Method: POST
• URL: https://api.mews.com/api/connector/v1/tasks/add

{
  "ClientToken": "{{$env.MEWS_CLIENT_TOKEN}}",
  "AccessToken": "{{$env.MEWS_ACCESS_TOKEN}}",
  "Name": "Guest Request: {{$json.task.category}}",
  "Description": "{{$json.task.description}} | Room {{$json.guest.room_number}} | Guest: {{$json.guest.first_name}}",
  "DepartmentId": "{{$env.MEWS_HOUSEKEEPING_DEPT_ID}}",
  "ServiceOrderId": "{{$json.guest.reservation_id}}"
}

Route C: Simple Notification (no dedicated housekeeping system)

Node 3C: Send to Microsoft Teams / Slack

• Channel: #housekeeping-requests

🏨 **New Guest Request**
Room: {{$json.guest.room_number}}
Guest: {{$json.guest.first_name}} {{$json.guest.last_name}}
Request: {{$json.task.description}}
Priority: {{$json.task.priority}}
Time: {{$json.timestamp}}

React with ✅ when completed.

Node 4: Log Task in Database (for tracking)

• Type: Postgres / Airtable / Google Sheets
• Action: Insert Row
• Fields: task_id, room, guest_name, category, status=pending, created_at, conversation_id

Node 5: Respond to AI Platform (acknowledge receipt)

• Type: Respond to Webhook
• Status: 200

{ "status": "accepted", "task_id": "generated_id", "estimated_minutes": 15 }

Task Completion Callback Workflow (separate n8n workflow)

• Trigger: Webhook from housekeeping system on task completion
• Path: /task-completed

{ "task_id": "xxx", "completed_by": "staff_name", "completed_at": "timestamp" }

• Method: POST
• URL: {{original_task.callback_url}}

{
  "event_type": "task_completed",
  "task_id": "xxx",
  "conversation_id": "conv_abc123",
  "completed_at": "2025-07-15T14:50:00Z"
}

This triggers the AI platform to send the guest a completion message: "Your extra towels have been delivered to room 412! Is there anything else I can help with? 😊"

Guest Messaging AI Disclosure and Privacy Compliance Module

1. AI DISCLOSURE (EU AI Act - mandatory transparency)

First-contact disclosure message (sent before or with first AI message)

# first-contact template and identity response

ai_disclosure_template:
  trigger: first_message_in_conversation
  position: prepended to first AI response OR sent as separate first message
  template: |
    💬 Welcome to {{property.name}} messaging!
    
    I'm an AI-powered virtual concierge here to help with your stay.
    A real person is always available — just reply HUMAN at any time.
    
    By continuing this conversation, you agree to our privacy policy:
    {{property.privacy_policy_url}}
    
    Reply STOP to opt out of messages.
  
  # Shorter version for returning guests (already disclosed):
  returning_guest_template: |
    Welcome back, {{guest.first_name}}! 😊 Your AI concierge is here.
    Reply HUMAN anytime for a real person, or STOP to opt out.

# When guest explicitly asks if they're talking to AI:
ai_identity_response: |
  Great question! I'm {{property.name}}'s AI-powered virtual concierge.
  I can help with most questions about your stay, amenities, and local area.
  For anything I can't handle, I'll connect you with our team right away.
  Just say HUMAN whenever you'd like to speak to a person. 🙂

2. OPT-IN / OPT-OUT HANDLING (GDPR + CCPA + TCPA)

opt_out_handling:
  trigger_keywords: ['STOP', 'UNSUBSCRIBE', 'OPT OUT', 'CANCEL', 'QUIT']
  action:
    - immediately_stop_all_automated_messages
    - send_confirmation: |
        You've been unsubscribed from {{property.name}} messages.
        You will no longer receive automated messages.
        Reply START to re-subscribe, or call us at {{property.phone}}.
    - update_guest_profile: opted_out_messaging = true
    - log_opt_out: timestamp, channel, guest_id
    - do_NOT_send_any_further_messages (including marketing, upsells, surveys)
    - exception: staff-initiated manual messages for critical stay information
      (e.g., safety alerts) are still permitted under legitimate interest

opt_in_handling:
  trigger_keywords: ['START', 'SUBSCRIBE', 'OPT IN']
  action:
    - update_guest_profile: opted_out_messaging = false
    - send_confirmation: |
        Welcome back! You'll now receive messages from {{property.name}}.
        Reply STOP anytime to unsubscribe.

CCPA Notice

For CCPA: Include 'Do Not Sell My Personal Information' link

ccpa_notice:
  applicable_when: guest.state == 'CA' OR property.state == 'CA'
  include_in: privacy_policy_url
  action_on_request: |
    If a guest requests data deletion or asks about data practices:
    1. Acknowledge the request
    2. Escalate to hotel management with CCPA/privacy tag
    3. Response: "I've forwarded your privacy request to our management team.
       They'll respond within 45 days as required. You can also email
       {{property.privacy_email}} directly."

3. DATA RETENTION POLICY

data_retention:
  conversation_messages:
    retention_period: 13_months  # Balance between operational need and minimization
    auto_delete: true
    legal_hold_override: true  # If flagged for legal, extend retention
  
  guest_profiles:
    retention_period: 36_months  # Align with typical loyalty program cycle
    anonymize_after: 36_months  # Don't delete — anonymize for analytics
    deletion_on_request: immediately  # GDPR right to erasure
  
  payment_data:
    retention: NEVER_STORE  # PCI DSS: Never store card data in messaging
    tokenized_references: 13_months
  
  analytics_data:
    retention: aggregated_indefinitely  # No PII in aggregated analytics
    individual_metrics: 13_months

4. PCI DSS 4.0.1 COMPLIANCE RULES

# card number detection and tokenized payment processing rules

pci_compliance:
  # Rule: NEVER collect, store, display, or repeat credit card numbers in messages
  card_number_detection:
    pattern: '\b(?:4[0-9]{12}(?:[0-9]{3})?|5[1-5][0-9]{14}|3[47][0-9]{13}|6(?:011|5[0-9]{2})[0-9]{12})\b'
    action_if_guest_sends_card_number:
      - DO NOT echo or confirm the number
      - immediately_respond: |
          ⚠️ For your security, please never share credit card numbers
          via text message. I've flagged this message for secure deletion.
          
          To make a payment, please use this secure link:
          {{generate_payment_link()}}
          
          Or visit the front desk where we can process your card safely.
      - flag_message_for_deletion: true
      - alert_security_team: true
      - log_pci_incident: true
  
  # Rule: Use only tokenized payment links
  payment_processing:
    method: tokenized_payment_links_only
    providers: [stripe_payment_links, adyen_pay_by_link, shift4_secure_link]
    link_expiration: 24_hours
    template: |
      Here's your secure payment link for {{charge_description}}:
      {{payment_link}}
      This link expires in 24 hours. Your card information is
      processed securely and never stored in our messaging system.

5. GDPR DATA SUBJECT REQUESTS

gdpr_requests:
  right_to_access:
    trigger_phrases: ['my data', 'what data do you have', 'data request', 'subject access']
    response: |
      I can help with your data request, {{guest.first_name}}.
      Under GDPR, you have the right to access your personal data.
      I'm connecting you with our data protection team who can
      process your request within 30 days.
    action: escalate_to_management(tag='GDPR_SAR')
  
  right_to_erasure:
    trigger_phrases: ['delete my data', 'forget me', 'erase my data', 'right to be forgotten']
    response: |
      I understand you'd like your data deleted. I've forwarded this
      request to our data protection team. They'll process your
      request and confirm within 30 days. You can also email
      {{property.dpo_email}} for direct assistance.
    action:
      - escalate_to_management(tag='GDPR_ERASURE')
      - create_task(department='management', priority='HIGH')

6. AUDIT LOGGING

audit_log:
  log_all:
    - ai_disclosure_presented: timestamp, guest_id, channel
    - consent_status_changes: timestamp, guest_id, old_status, new_status
    - opt_out_requests: timestamp, guest_id, channel, keyword_used
    - data_subject_requests: timestamp, guest_id, request_type
    - pci_incidents: timestamp, guest_id, type, action_taken
    - escalation_events: timestamp, guest_id, reason, assigned_to
    - message_deletion_events: timestamp, reason, authorized_by
  storage: platform_native + export_to_msp_siem
  retention: 24_months_minimum

Testing & Validation

• NETWORK TEST: From a staff tablet on VLAN 200 (Staff-Ops), confirm you can access the AI messaging platform dashboard (https://app.akia.com) with <2 second page load time. Verify that the same tablet CANNOT access devices on VLAN 300 (PMS) directly via IP.
• NETWORK TEST: From a device on VLAN 100 (Guest WiFi), attempt to access a device on VLAN 200 or VLAN 300. Confirm the connection is BLOCKED by the firewall. This validates PCI DSS network segmentation.
• SMS DELIVERY TEST: Send a test SMS to 3 different carrier numbers (AT&T, Verizon, T-Mobile) from the provisioned Twilio number. Verify delivery within 5 seconds. Reply from each number and verify the AI agent responds appropriately within 10 seconds.
• WEBCHAT TEST: Visit the hotel website in Chrome, Firefox, Safari, and on iOS/Android mobile browsers. Verify the webchat widget appears within 5 seconds, is clickable, and allows message input. Send a test message and verify AI response within 5 seconds.
• PMS SYNC TEST: Create a test reservation in the PMS with check-in date = tomorrow. Verify that within 15 minutes, the reservation appears in the AI messaging platform dashboard with correct guest name, dates, room type, and contact information.
• AI ACCURACY TEST - INFORMATIONAL: Send 20 different informational questions covering all knowledge base categories (e.g., 'What time is checkout?', 'Do you allow pets?', 'Where's the nearest pharmacy?', 'What's the WiFi password?'). Score each response for accuracy and completeness. Target: 90%+ correct answers with no fabricated information.
• AI ACCURACY TEST - SERVICE REQUESTS: Send 5 housekeeping requests ('Can I get extra towels?', 'Please clean my room', 'I need more shampoo', 'Can I get a rollaway bed?', 'Turndown service please'). Verify each creates a task in the housekeeping system with correct room number and request details.
• ESCALATION TEST: Send messages that should trigger escalation: 'I want to speak to a manager', 'My bill is wrong', 'I feel unsafe in my room', 'This is completely unacceptable'. Verify each is escalated to the human inbox within 30 seconds with correct priority level.
• AI DISCLOSURE TEST: Start a brand new conversation (new phone number). Verify the AI disclosure message is sent within or before the first AI response, clearly stating it is an AI system with a link to the privacy policy and opt-out instructions.
• OPT-OUT TEST: Send 'STOP' to the messaging number. Verify immediate opt-out confirmation is received. Then trigger a pre-arrival workflow for the opted-out guest and verify NO message is sent. Send 'START' to re-subscribe and verify confirmation.
• PCI COMPLIANCE TEST: Send a fake credit card number (use Stripe test number 4242424242424242) in a message. Verify the AI does NOT echo the number, responds with a security warning, generates a secure payment link instead, and the incident is logged.
• UPSELL WORKFLOW TEST: Create a test reservation with check-in in 2 days. Verify the pre-arrival welcome message is sent at the configured time. Respond 'YES' to the early check-in offer and verify a payment link is generated and the reservation is updated in the PMS.
• MULTI-LANGUAGE TEST (if applicable): Send messages in Spanish, French, and Mandarin. Verify the AI responds in the guest's language or gracefully escalates if the platform doesn't support that language.
• FAILOVER TEST: Temporarily block outbound HTTPS (port 443) on the firewall to simulate internet outage. Verify that queued messages are delivered once connectivity is restored, and that the MSP monitoring system generates an alert within 5 minutes.
• STAFF TABLET TEST: On each staff tablet, verify: (a) Akia app is installed and logged in, (b) push notifications are enabled and audible, (c) staff can pick up an escalated conversation and respond, (d) staff can close a resolved conversation, (e) MDM enrollment is active and device appears in Intune.
• LOAD TEST: Using the Twilio API, send 50 concurrent test messages to the AI platform and verify all receive responses within 30 seconds with no dropped messages. This simulates a busy check-in period at a 120-room property.
• AFTER-HOURS TEST: Send a guest request at 2:00 AM (outside front desk hours). Verify the AI handles the informational request autonomously. Then send an escalation-required message and verify it routes to the on-call duty manager's notification channel.
• END-TO-END GUEST JOURNEY TEST: Simulate a complete guest lifecycle: (1) receive pre-arrival welcome 48 hours before check-in, (2) complete digital check-in via link, (3) send in-stay request after check-in, (4) receive mid-stay satisfaction check, (5) receive late checkout offer before departure, (6) receive post-stay review solicitation after checkout. Verify each touchpoint fires correctly with accurate personalization.

Client Handoff

Conduct a formal 90-minute handoff session with the hotel General Manager, Front Desk Manager, and all staff who will use the escalation inbox. Cover the following topics in order:

Documentation to Leave Behind

• Printed and laminated quick-reference cards for each staff station (escalation inbox workflow, 5 key actions)
• Staff training video recording (record the dashboard training portion)
• Knowledge base update request form/process document
• Compliance quick-reference card (PCI, GDPR, opt-out handling)
• Network diagram with VLAN layout, IP addresses, and credentials location
• Vendor contact information (Akia/Canary support, Twilio support)
• Emergency procedure document (system outage, what to do if AI gives wrong info)
• Monthly reporting template with KPI targets
• Full system architecture document with all integration endpoints, API keys (referenced by password manager entry, not plaintext), and webhook URLs

Success Criteria to Review Together

Maintenance

Monthly MSP Responsibilities:

Model Retraining / Knowledge Base Refresh Triggers

• Hotel renovates or adds new amenities
• Restaurant menu or hours change
• Seasonal transitions (summer/winter pool hours, holiday schedules)
• New local attractions or events
• Policy changes (cancellation, pet, parking)
• AI accuracy drops below 80% for two consecutive weeks
• New complaint category emerges (>5 similar complaints in a month)
• Guest feedback identifies consistent AI error pattern

SLA Considerations

• Platform Uptime: 99.9% (vendor SLA) — MSP monitors and escalates to vendor
• Network Uptime: 99.5% (MSP responsibility) — redundant ISP recommended
• Knowledge Base Updates: 24-hour turnaround for standard changes, 4-hour for critical corrections
• Emergency Support: 1-hour MSP response for system-down scenarios
• Staff Support: 4-hour response for non-critical operational questions

Escalation Path

Alternatives

Canary Technologies (Mid-Market/Enterprise Alternative)

Replace Akia with Canary Technologies as the primary AI messaging platform. Canary is the market leader with a $600M valuation, 25,000+ hotel customers, and the broadest feature set including AI Voice (handles phone calls), AI Guest Messaging (SMS/email/webchat), Dynamic Upsells, Contactless Check-in, and white-label branding. Canary offers a single platform that combines text and voice AI, eliminating the need for separate voice solutions.

• Recommend Canary when: client has 150+ rooms, needs voice AI, is part of a hotel group, or has enterprise PMS (OPERA).
• Recommend Akia when: client has <150 rooms, needs fastest deployment, is budget-conscious, or values simplicity.

HiJiffy (International/European Properties)

Replace Akia with HiJiffy for properties with significant international guest populations or European locations. HiJiffy is headquartered in Europe (GDPR-native), supports 130+ languages via its proprietary Aplysia AI engine, and serves 1,800+ hotels globally. Strong WhatsApp and Booking.com native integrations make it ideal for properties heavily reliant on OTA bookings and international travelers.

• Recommend HiJiffy when: client is European-based, has >30% international guests, relies heavily on Booking.com/Expedia, or has strict GDPR requirements.
• Not recommended when: property is US-only with primarily domestic guests and uses OPERA or Cloudbeds (where Akia/Canary have stronger integrations).

Custom Build with Botpress + Twilio + OpenAI

Instead of a purpose-built hospitality platform, build a custom AI messaging agent using Botpress Cloud (visual chatbot builder), Twilio (messaging infrastructure), and OpenAI GPT-5.4 (LLM inference). The MSP develops and maintains the entire solution, including PMS integration via custom API code, conversation flows, knowledge base, and analytics.

Tradeoffs

• Lower monthly SaaS cost ($89/month Botpress + ~$50/month Twilio + ~$30/month OpenAI ≈ $170/month vs. $360–$1,200/month for SaaS platforms) but significantly higher MSP labor for ongoing maintenance (10+ hours/month vs. 2–4 hours/month).
• No pre-built PMS integrations—must build and maintain custom API connections.
• No hospitality-specific features (upsell engine, digital check-in, payment links) unless custom-built.
• No vendor support—MSP owns all troubleshooting.

Runnr.ai (Ultra-Budget European Option)

For extremely budget-conscious properties or small European hotels (under 50 rooms), Runnr.ai offers AI guest messaging automation at approximately €0.75 per room per month. Focuses primarily on WhatsApp and Booking.com inbox automation. Claims 95% automation rate and 45-minute setup time. Direct integration with OPERA Cloud, Mews, and StayNTouch.

Recommend when: client is a small European hotel (<50 rooms), budget is extremely tight, primary guest channel is WhatsApp/Booking.com, and property needs basic automation rather than comprehensive guest experience management.

Conduit (Voice + Text Combined)

For properties where phone call volume is a major pain point (e.g., resorts, properties with older demographics, high-service luxury hotels), Conduit (formerly HostAI) is the only platform that handles both voice calls and text-based messaging through a single AI system. The AI answers phone calls, understands spoken requests, and handles them autonomously or transfers to staff.

Tradeoffs

Starting at $500/month for self-serve (up to $2,000/month for mid-sized properties with full service), Conduit is priced comparably to Canary. The unique voice AI capability eliminates the need for a separate IVR system or phone answering service. However, voice AI is newer technology with higher error rates than text—expect 60–70% autonomous resolution on voice vs. 80%+ on text. Implementation takes 2–4 weeks including voice AI training and PMS integration.