Three weeks to deploy a 3D hologram capsule in a hotel lobby
How we go from brief to a working AI greeter installation — avatar selection, voice integration, on-site commissioning.
A boutique European hotel group wanted an AI greeter capsule in the lobby of three flagship properties. They had a brand voice in mind. They did not have an LLM, an avatar, or a TTS pipeline. We delivered three working installations in three weeks.
Week 1 — brief and avatar
Kickoff call established the brand voice (warm, four-language: EN/FR/DE/IT), the use cases (check-in, wayfinding, restaurant booking), and the integration scope (PMS lookup for guest names, no PCI). We delivered three avatar concepts in 48 hours; the brand picked one. Avatar rigging and lip-sync mapping took the rest of week 1.
Week 2 — integration
We integrated their existing OpenAI Enterprise account for intent classification and ElevenLabs for voice synthesis. PMS lookup ran over OAuth to their Mews integration. Avatar response time landed at 800 ms (target was 1.2 s). Stress-test ran for 72 hours continuous operation in our QC room.
Week 3 — install
Three WT-HC75 capsules shipped FOB Dongguan to Paris, Berlin, and Milan. On-site commissioning was 4 hours per property — calibrating the IR microphone array for lobby acoustics is the longest step. Brand sign-off was on the third day.
Two months in
Engagement rate (guest walks up + interacts) sits at 38% of lobby foot traffic. Average session is 90 seconds. The two most-asked questions are restaurant hours and gym access. The hotel group is rolling out four more properties.