← → or scroll to navigate
Sponsorship Proposal · Google Cloud for Startups

StayPilot

The AI front desk & hotel management platform for Vietnam's hospitality market — built on Google Cloud.

Gemini 2.5 Flash Live in production GCE asia-east1
StayPilot × Google Cloud
01 · The Problem

Small & mid-size hotels are bleeding revenue every day

💬

Guests message 24/7, front desk works 8 hours

Bookings arrive via Zalo / Facebook all night. A few minutes of silence — the guest books elsewhere.

💸

15–25% OTA commission

OTA dependency erodes the already-thin margins of 10–200 room hotels.

📒

Manual operations

Room checks in group chats, verbal price quotes, Excel ledgers — errors and lost guests.

🏷️

International PMS is overkill

Foreign solutions cost too much, don't speak Zalo, and don't fit Vietnam's night/day-rate business.

Hospitality pain points02 / 10
02 · The Solution

StayPilot — the AI room-sales agent that never sleeps

An AI chatbot that answers guests on Zalo & Facebook Messenger in natural Vietnamese: quotes prices, checks availability, closes bookings, verifies bank transfers — and syncs everything back to the hotelier's management console.

24/7

Running in real production at Bonita Hotel (2 properties) since March 2026 — not a demo.

< 1′

Replies to guests in under a minute, at any hour — including holidays and Tết.

100%

Built on Google Cloud infrastructure: Compute Engine + Gemini, ready to scale onto Vertex AI.

Solution03 / 10
03 · Features

AI that closes the full sales loop — not just chat

🤖

Consult & quote

Gemini understands colloquial Vietnamese; sends rate cards and room photos in context.

📅

Availability & booking

Checks vacancies, blocks past dates, auto-computes night / day / multi-day pricing.

🧾

Vision payment verification

AI reads bank-transfer screenshots, rejects fake or wrong ones before staff confirmation.

🎛️

Realtime admin control

Pause / resume the bot per guest; staff take over any conversation live from the web dashboard.

💎

Loyalty & upsell

Recognizes returning guests, offers room upgrades, grows order value.

📊

Structured data store

Guests, bookings & conversations persisted in PostgreSQL (Drizzle ORM) — queryable, reportable, no spreadsheets.

System features04 / 10
03 · Features — PMS roadmap

From chatbot to multi-hotel PMS (SaaS)

Deploy-per-hotel architecture: every property runs its own isolated instance and database — zero data mixing, per-hotel branding out of the box.

🛎️

Front desk

Calendar-style reservation board, check-in / check-out, room assignment.

🧹

Housekeeping

Realtime room status, task assignment for cleaning staff.

💰

Rates & billing

Seasonal / hourly rate engine, invoicing, payment reconciliation.

📈

Reporting

Occupancy, RevPAR, guest sources — one screen.

🔗

OTA integration

Two-way inventory sync, less overbooking.

🏢

Instance-per-hotel

One container + one PostgreSQL per property — hard data isolation by design.

Product roadmap05 / 10
04 · Benefits

Measurable value for hotel owners

0 missed

No guest left waiting: the bot covers the night shift; staff handle exceptions only.

−OTA

More direct bookings via Zalo/Facebook — every order shifted from OTA saves 15–25% commission.

~$40/mo

Full per-hotel infrastructure cost — one dedicated VM on GCP, measured from real billing, not a theoretical estimate.

1 source

Centralized guest & booking data — the foundation for loyalty, remarketing, and accurate reporting.

Benefits delivered06 / 10
05 · Architecture — live today

In production today, on Google Cloud

Guest channels

Zalo OA

chat + payment photos

FB Messenger

webhook

StayPilot Core

Compute Engine

asia-east1 · Bun + auto-restart launcher
AI brain · booking engine

Web dashboard

live chats · staff reply bridge

PostgreSQL 17

Drizzle ORM · isolated per hotel

AI layer — 3-tier fallback

1 · Gemini AI Studio

gemini-2.5-flash

2 · Vertex AI

asia-southeast1, ADC

3 · OpenRouter

independent-quota backstop

Quota-fault-tolerant by design: when one AI tier hits a 429 limit, the system fails over to the next — guests never see the bot freeze.

Current architecture07 / 10
06 · Deployment

Scaling roadmap on GCP

✓ Today

Production foundation

GCE e2-medium (asia-east1) · Gemini API · one real hotel, two properties in operation.

Phase 1 · Q3/26

Scale the fleet, hotel by hotel

Docker-ready template — one Compute Engine VM + isolated PostgreSQL per hotel · Secret Manager · 5–10 hotel pilot.

Phase 2 · Q4/26

Standardize AI & analytics

Vertex AI as the primary tier · BigQuery + Looker Studio owner reports · Cloud Monitoring.

Phase 3 · 2027

Regional scale

1,000+ hotels across Vietnam → Southeast Asia · multi-region VM fleet with managed instance groups.

GCP deployment roadmap08 / 10
07 · The Ask

We need GCP credits to accelerate

🧠

Vertex AI / Gemini

Tokens for a 10-hotel pilot — promoting the primary AI tier from AI Studio to enterprise Vertex AI.

☁️

Multi-tenant infrastructure

Compute Engine, Cloud SQL, Secret Manager, Storage — building the SaaS that serves many hotels.

📊

Analytics

BigQuery + Looker Studio: occupancy and revenue reporting for hotel owners.

In return: a Vietnamese hospitality SaaS case study running 100% on Google Cloud, with real paying customers — not an idea on paper.

Sponsorship ask09 / 10
Thank you

Let's bring Vietnam's hotels
onto Google Cloud.

StayPilot — the AI front desk for hospitality.

StayPilot × Google Cloud