Skip to content
AI Otel

The platform

Intelligence with a hotel underneath it.

Most AI in hospitality is a chatbot bolted onto old software. AI Otel is the reverse: a complete operating system whose every function was built to be run by intelligence, human or machine.

The architecture

One contract. Three consumers.

Every function is defined once, in one API contract, and served to three consumers from a single source of truth. Nothing is bolted on for the machines; the machines are first-class.

  • /rpcHumans

    The operator interface: dashboards, pricing, guests, reports.

  • /apiPartners

    Channel managers, payment providers, fiscal integrations.

  • /mcpAI agents

    The live agent endpoint: discovery, booking, invoicing, audit.

The AI

Three levels of brain. One promise: a hotel's knowledge stays its own.

  1. Industry AI

    Learns across all hotels: demand patterns, seasonality, market behavior no single property could ever see alone.

  2. Platform AI

    Carries AI Otel's hospitality know-how, revenue practice, compliance handling, to every hotel on the platform.

  3. Hotel AI

    Private to each hotel. Trained on that property alone, and it cannot be copied or transferred to anyone else.

How it is trained

Built from strong open models, Llama, Qwen, Gemma, and Turkish models: Trendyol, Kumru, T3AI. Fine-tuned on hospitality data, grounded in each hotel's live systems, and retrained continuously.

Privacy is architectural, not contractual: federated learning means a hotel's private data never leaves the hotel.

Turkish models, trained in Türkiye.

The brain is architected and entering training, a deliberate sequence: the platform first, the intelligence on top.

aiotel-mcp · session replay
  1. agent connected · scoped credentials · audit trail on
  2. tools discovered: availability, pricing, booking, invoicing
  3. query: 2 nights, 2 guests, sea view, late check-out
  4. availability confirmed · rate returned with taxes itemized
  5. booking created · confirmation issued to guest
  6. e-Arşiv invoice generated · legally valid in Türkiye
  7. session closed · full record kept for the hotel
Replay of a representative session against the live MCP endpoint, where AI agents complete real bookings end to end.

In production today

Working AI, before the brain.

The three-level brain is entering training. These capabilities already run in production.

  • Conversational agent

    Guest and operator conversations with real tool execution behind them.

  • Voice agent

    Speaks and understands five languages, around the clock.

  • Pricing engine

    Five factors weighed continuously: demand, season, market, events, inventory.

  • Autopilot

    Acts on its own within limits, and asks when a change is material.

  • Anomaly detection

    Catches the strange booking, the odd rate, the broken feed, early.

Deployment

The model follows the hotel’s requirements. Not the other way around.

  • Global

    For hotels without data-residency requirements: the strongest available frontier models, chosen for capability.

  • Türkiye-resident

    Strong open models, Llama, Qwen, Gemma, self-hosted on infrastructure in Türkiye. Data never leaves the country.

  • Sovereign

    Turkish models, Trendyol, Kumru, T3AI, trained and served in Türkiye, for hotels and institutions that require a fully domestic stack.

One platform, three deployment postures. Every hotel chooses where its intelligence lives.

Security & isolation

Engineered like the regulated system it is.

  • Tenant isolation

    Row-level security on every query path; a build fails if isolation regresses.

  • Per-hotel encryption

    Each hotel’s data is encrypted with its own keys, envelope-managed.

  • Hash-chained audit

    Every action, human or agent, lands in a tamper-evident audit trail.

Training methods, named plainly

  • LoRA/QLoRA
  • RAG
  • SFT+DPO
  • Federated learning

The platform

What we have built

Six capabilities, one system. Stated plainly, with honest status.

  • Live

    AI-native PMS

    The hotel operating layer: pricing, channels, guests, reports, compliance.

  • Built · Türkiye rollout

    Guest app & tourism routing

    Verified local experiences, booked and paid in-app, value kept in Türkiye.

  • Architected · entering training

    Three-level AI brain

    Industry, platform, and per-hotel intelligence with federated privacy.

  • Live

    Agent (MCP) endpoint

    AI agents book end to end with scoped credentials and full audit records.

  • In development

    Agency hub

    Connects Turkish travel agencies to hotel inventory with smarter pricing.

  • Live

    Native compliance

    KBS, KVKK, e-Fatura, e-Arşiv, ÖKC, Mali Mühür, built into the data model.

See a booking happen with no human in the loop.

Request a demo and we will show you the product live, including an AI agent completing a real booking end to end, with a valid Turkish invoice at the end of it.