SparkDesk/AI receptionist for hotels
Your 24/7 intelligent hotel receptionist.
Room bookings, spa reservations, dinner tables. Voice or text, in 57 languages. Your guests get a real conversation. Your team gets the night off.
PMS IntegrationVoice & TextGDPR Compliant
Always at the desk · 24/701The AI Receptionist
Everything your front desk handles. In a conversation, not a phone tree.
Core Capabilities
sparkdesk handles every step a guest asks for, from the first message to checkout, by voice or text.
Room Accommodations
- Real-time availability checks
- Complete booking creation
- Intelligent upselling
- Modifications & cancellations
- Pre-arrival check-in
Spa & Wellness
- Service catalog navigation
- Therapist scheduling
- Preference handling
- Package bookings
- Health questionnaires
Restaurant & Dining
- Table management
- Menu & ambience info
- Dietary restrictions
- Special occasions
- Multi-venue support
Conversational Intelligence
Instead of an IVR menu that times out, a voice that listens, follows context, and answers like a person.
Natural conversational flow
sparkdesk responds at human pace, with the small filler words people actually use ('let me check that for you', 'one moment') while it queries your systems. No dead air on the line.
Interruption handling (barge-in)
Guests can cut the AI off mid-sentence. It stops, takes the new instruction, and picks up from there, the way a person at the desk would.
Tone detection and sentiment
If a guest sounds frustrated, the AI softens its tone. If something feels off, it pushes the call to your staff with the context already in hand.
57 languages, accents, idioms
57 languages, switchable mid-conversation. Accents, regional phrasing, and informal booking talk ('a quiet night, second weekend of June') all land cleanly.
Warm handoffs to your staff
When the AI can't close a request, it warm-transfers to your team with the guest's intent and the details already gathered. Nobody has to repeat themselves.
02Real-World Scenarios
See sparkdesk in action.
Four real conversations, from a three-line text request to a multi-room weekend booking with spa and dinner.
I'd like to book a weekend stay for my anniversary next month, from the 12th to the 14th.
I'd be happy to help you plan your anniversary getaway! For the dates of Friday, the 12th to Sunday, the 14th, we have our Deluxe King and our Oceanview Suite available. Which would you prefer?
Oceanview suite, please. Also, can I book a couples massage for Saturday afternoon and dinner at your steakhouse that evening?
Excellent choice. I have reserved the Oceanview Suite. For the spa, we have a 60-minute couples massage available at 2:00 PM or 4:30 PM on Saturday. For dinner at The Prime Cut, I can reserve a romantic window table for two at 7:30 PM. Does this itinerary work for you?
The 2:00 PM massage and the 7:30 dinner sound perfect. Let's go with that.
Wonderful! Everything is confirmed. Your anniversary itinerary: Oceanview Suite, Friday the 12th to Sunday the 14th. Couples massage at 2:00 PM Saturday. Dinner for two at The Prime Cut, window table, 7:30 PM Saturday. A confirmation email is on its way. Happy anniversary!
03Plugs into your stack
Integrates with the systems you already run.
sparkdesk plugs into the PMS, phones, and payment stack you already run.
Core property management
Deeply integrated with Mews.
Live sync for room inventory, reservations, guest profiles, and central billing.
- Room inventory
- Reservations
- Guest profiles
- Central billing
- Folio adjustments
Essential infrastructure
Telephony and voice
Sits on top of the PBX you already have, so the voice quality lands without replacing a single phone.
Secure payment links
During the call, SparkDesk texts a one-time Stripe link. The guest pays on their own phone. Card details never touch the conversation.
Expanding ecosystem
On the roadmap
04Enterprise security
Built for the audit your buyers will run.
- Zero-knowledge payments
- Payment details never reach the voice channel. The guest gets a one-time link by SMS and pays on their own device. The AI never sees a card number.
- Transcript masking
- Card numbers and personal identifiers are masked at capture, so nothing sensitive ever lands in a transcript or log.
- Verified modifications
- Set the verification bar you want, from Caller ID match to full two-factor, before any change to a booking goes through.
- Global privacy standards
- Built for GDPR, CCPA, and the local regulations behind them. Retention windows are configurable per property, consent is explicit.
- Full encryption at rest
- Guest records, conversation logs, and transaction metadata are encrypted at rest. A stolen disk reveals nothing readable.
05About SparkDesk
Built in Switzerland for hotels that take service seriously.
sparkdesk was built in Zurich for properties that treat the front desk as a craft, not a queue. The technology is current AI; the brief is older than that.
AI engineering
Current-generation language models with tone detection and pacing that don't make a guest hang up.
Built for hotel operations
Designed around how hotels actually run: shift changes, walk-ins, late checkouts, the lot.
Security and compliance
GDPR, PCI-DSS, Swiss FADP. The data hygiene a procurement team will check, before they ask.
The team
sparkdesk is the first product from Polygon AI, a Swiss company building software for high-availability operations. The founders, Slawomir Zak and Lena Sophia Bolliger, pair a long career in cloud infrastructure with current research on how language models can hold a conversation. They built sparkdesk because hotels need both.
Slawomir Zak
CEO, Polygon AI
25 years in systems engineering and DevOps, eleven of them at Google running operations and capacity planning. Co-founder of Billin, a SaaS subscription billing platform. Slawomir designs the parts of sparkdesk that have to stay up at 3am: the integrations into the PMS, the telephony bridge, the queues that run when nobody is watching.
Lena Sophia Bolliger
CTO, Polygon AI
AI and NLP researcher, Ph.D. candidate at the University of Zurich. Lena works on fine-tuning large language models and aligning their behavior with how people actually think and talk. She designs the part of sparkdesk a guest experiences directly: how it follows context, when it pauses, when it asks again, when it picks up the phrasing the guest just used.
06Get in Touch
Want a demo on your own property?
Email us with a few details about your property. We'll set up a demo on your PMS, your phone setup, and a scenario from your guest book.