The CallSphere Hotel Stack: 11 Agents That Run a Property End-to-End

TL;DR

CallSphere's hotel platform ships 11 specialist AI agents built on OpenAI Agents SDK + Realtime API. They handle every guest-facing conversation — reservations, check-in, housekeeping, group sales, night audit — with deep PMS integration.

Why 11 Agents, Not One

A single monolithic agent "doing everything" collapses under load. It hallucinates rate quotes, forgets cancellation policies, and can't cleanly escalate. CallSphere uses hierarchical handoffs between narrow specialists, each with its own tools and guardrails.

flowchart LR
    CALLER(["Guest or Prospect"])
    subgraph TEL["Telephony"]
        SIP["Twilio SIP and PSTN"]
    end
    subgraph BRAIN["Hotel Concierge AI Agent"]
        STT["Streaming STT<br/>Deepgram or Whisper"]
        NLU{"Intent and<br/>Entity Extraction"}
        TOOLS["Tool Calls"]
        TTS["Streaming TTS<br/>ElevenLabs or Rime"]
    end
    subgraph DATA["Live Data Plane"]
        CRM[("CRM and Notes")]
        CAL[("Calendar and<br/>Schedule")]
        KB[("Knowledge Base<br/>and Policies")]
    end
    subgraph OUT["Outcomes"]
        O1(["Reservation confirmed"])
        O2(["Room service order"])
        O3(["Front desk handoff"])
    end
    CALLER --> SIP --> STT --> NLU
    NLU -->|Lookup| TOOLS
    TOOLS <--> CRM
    TOOLS <--> CAL
    TOOLS <--> KB
    NLU --> TTS --> SIP --> CALLER
    NLU -->|Resolved| O1
    NLU -->|Schedule| O2
    NLU -->|Escalate| O3
    style CALLER fill:#f1f5f9,stroke:#64748b,color:#0f172a
    style NLU fill:#4f46e5,stroke:#4338ca,color:#fff
    style O1 fill:#059669,stroke:#047857,color:#fff
    style O2 fill:#0ea5e9,stroke:#0369a1,color:#fff
    style O3 fill:#f59e0b,stroke:#d97706,color:#1f2937

Agent 1: Concierge (Triage)

The front door. Identifies the caller by phone number, pulls their guest profile, detects intent, and routes to the right specialist. Handles basic questions (hours, location, amenities) without handoff.

Tools: lookup_guest_by_phone, detect_intent, get_hotel_info, route_to_agent

Agent 2: Reservation

Handles new bookings. Searches availability, quotes best-available-rate, enforces rate parity, collects guest details, processes deposits via Stripe/Square, writes to PMS.

Tools: search_availability, quote_rate, collect_guest_details, process_deposit, create_reservation

Agent 3: Check-In

Express check-in for existing reservations. Verifies ID, captures incidentals authorization, issues mobile keys via Salto/Assa Abloy, confirms room assignment.

Tools: verify_reservation, capture_incidentals, issue_mobile_key, assign_room

Agent 4: Check-Out

Folio review, charge disputes, payment capture, receipt delivery, loyalty point posting.

Tools: pull_folio, dispute_charge, capture_payment, email_receipt, post_loyalty_points

Agent 5: Housekeeping

Takes voice reports from housekeeping staff ("room 214 clean"), updates PMS room status, creates maintenance tickets for issues ("AC broken in 308").

Tools: update_room_status, create_maintenance_ticket, assign_cleaner, report_inspection

Agent 6: Guest Services

Amenity requests, local recommendations, restaurant reservations, wake-up calls, extra towels, late check-out.

Tools: create_service_request, schedule_wake_up, book_restaurant, extend_check_out

Agent 7: Group Sales

Qualifies RFPs for room blocks, events, weddings. Checks block inventory, drafts proposal outlines, hands warm leads to DOSM with full transcript.

Tools: check_block_inventory, draft_proposal, schedule_site_visit, notify_dosm

Agent 8: Revenue Signals

Not guest-facing. Listens to inbound call patterns, detects demand signals, flags rate parity issues, alerts revenue manager to opportunity.

Tools: analyze_demand, check_parity, alert_revenue_manager

Agent 9: OTA Channel

Syncs rates and inventory across OTAs via Siteminder/Cloudbeds. Handles channel conflicts and restriction updates.

Tools: push_rates, sync_inventory, handle_overbooking

Agent 10: Loyalty

Recognizes VIPs on inbound calls, applies loyalty discounts, triggers room upgrades, logs preferences for future stays.

Tools: lookup_loyalty, apply_discount, trigger_upgrade, log_preference

Agent 11: Night Audit + Emergency

Runs 12 AM–7 AM. Handles late arrivals, issues keys, verifies reservations, and escalates emergencies via a ladder of call + SMS until acknowledged.

Tools: late_check_in, verify_identity, classify_emergency, escalate_to_on_call

How They Work Together

A guest calls at 2 AM asking about a wedding block for next June. Flow:

1. Concierge identifies intent: group sales
2. Hands off to Group Sales Agent with full context
3. Group Sales checks block inventory, drafts proposal
4. Notifies DOSM by email with transcript + next steps
5. After hours, hand back to Night Audit for immediate booking needs

All 11 agents share the same guest profile, PMS state, and conversation history.

FAQ

Q: Do all 11 agents come standard? A: Yes, on the Orchestrate and Custom Build plans. Act includes 5 core agents (Concierge, Reservation, Check-In, Check-Out, Night Audit).

Q: Can I customize the agents? A: Yes. Policies, tone, escalation rules, and upsell logic are all configurable per property.

Q: How does this compare to building it yourself on the OpenAI API? A: Building 11 agents with proper handoffs, PMS integration, multilingual support, and guardrails is a 6–12 month engineering project. CallSphere ships it turnkey in 3–7 days.

---

Related: Hotel industry | Compare CallSphere vs ASI PMS

#CallSphere #HotelAI #MultiAgent #AgenticAI #Hospitality