Fontainebleau Miami Beach · Leadership AI Workshop
Session 1 of 2 · 90 minutes · AM & PM cohorts
All data on this page is simulated for training — no real guest, financial, or hotel records

AI on the Floor — A Live Demo for Fontainebleau Leaders

A working hotel dashboard and AI assistant, built with the same everyday tools you already own — Microsoft 365, Copilot, and data from systems like Opera. Move the levers and watch the property respond.

Scenario
Levers
Data guardrails
Presets
Occupancy · 12-mo forecast
average across horizon
RevPAR · forecast
ADR × occupancy
Guest response time
routine inquiries, AI-assisted
Leader hours recovered
per month, across ~60 leaders
Governance readiness
composite of 3 risk areas
01

See the next 12 months, not just the last 12

24 months of occupancy history → 12-month AI forecast with confidence band

Predictive demand

Nightly occupancy — history & forecast

AI applied · Time-series forecasting horizon: 12 mo · 1,594 rooms
Historical actuals AI forecast 90% confidence band

Staffing the forecast

Where the model shifts the roster

AI applied · Optimization model

Forecast-driven scheduling vs. today's static rosters, by department (scheduled FTEs for the peak period).

Current roster AI-optimized
02

Turn the forecast into revenue

Dynamic pricing + upsell uplift over 12 months — and the reasoning behind it, in plain English

Financial model

Cumulative net uplift — 12-month horizon

AI applied · Dynamic pricing
Cumulative net uplift Investment recovered zone

Audit the model

"Explain the variance"

AI applied · Explainable AI

A strong leader doesn't just accept the number — they interrogate it. Ask the model why the forecast moves:

What is driving this forecast?
    03

    The guardrails in action — try the AI assistant

    Three real requests a manager might make. Flip the Data guardrails toggle above and try the red one again.

    FB
    Fontainebleau AI Assistant
    Internal pilot · Microsoft 365 tenant
    Guardrails on
    Pick a request below to see how the assistant — and the guardrails — respond.
    Try a request

    Know your data

    The 30-second classification check

    Before anything goes into an AI tool, every leader asks one question: what color is this data?

    Green — Public

    Pool hours, restaurant menus, event calendars, published rates. Any approved AI tool is fine.

    Amber — Internal

    Room numbers, staffing plans, SOPs, unpublished occupancy. Only inside the Fontainebleau M365 tenant (Copilot) — never public chatbots.

    Red — Confidential

    Guest identity + PII, payment cards (PCI), health notes, contracts. Never enters any AI tool. No exceptions.

    04

    Govern the risk before it governs you

    Risk exposure updates as adoption scales — toggle the guardrails above and watch

    Guest privacy & PCI

    Guest identity, preferences, and payment data are the crown jewels — and a legal and brand exposure if they leak into public AI tools. Mitigation: the red-amber-green classification every leader learns today.

    Accuracy & brand voice

    An AI that confidently invents a spa policy or comps a suite is a guest-recovery incident waiting to happen. Mitigation: human review on every guest-facing message — AI drafts, the manager decides.

    Shadow AI

    If approved tools are slow or missing, staff will quietly use personal chatbots — invisible and ungoverned. Mitigation: give every department a sanctioned, capable alternative and a clear usage policy.

    05

    A manager's AI day — with the stack you already own

    No new systems required: Opera stays the system of record, Microsoft 365 is the safe workspace

    8:00 AM · Brief
    Microsoft Copilot 365

    Summarizes overnight Opera reports, arrivals, VIPs, and yesterday's guest feedback into a one-page morning brief — inside your tenant, so internal data stays internal.

    Today's skillPrompting for daily operations
    Midday · Respond
    AI drafting + human review

    Drafts guest recovery notes, shift handovers, and interdepartmental updates in seconds. The manager edits tone and approves — AI never sends on its own.

    Today's skillThe draft–review–send habit
    4:00 PM · Plan
    Predictive dashboard

    Occupancy forecast, staffing shifts, and pricing signals — the dashboard you're looking at right now, built with AI in an afternoon, not a six-month IT project.

    Today's skillReading & challenging a forecast
    Always on · Protect
    The guardrails

    The red-amber-green data check, the approved-tools list, and human sign-off on anything guest-facing. Governance is what lets the other three run at full speed.

    Today's skillThe 30-second classification check
    06

    We practiced what we preach — how AI built this session

    Every material you've seen today was AI-made under the same guardrails you just learned: classified data, approved tools, human review

    Step 1 · Listen & summarize
    The planning meeting

    The video call with Michelle and Ricardo was transcribed and summarized by AI into a structured brief — goals, next steps, room setup, tech stack — in minutes, not an afternoon of note-typing.

    Guardrail appliedAmber data — meeting notes stayed in approved tools; no guest data involved
    Step 2 · Analyze & recommend
    The Top-8 opportunity map

    AI cross-referenced Fontainebleau's management roles with hotel-industry AI use cases and produced the eight opportunities this workshop is built around — each mapped to the departments in this room.

    Guardrail appliedPublic + internal context only; recommendations reviewed by AU faculty
    Step 3 · Design & build
    This dashboard

    Claude built this entire interactive page — the forecast model, the charts, the AI assistant, the guardrails — from a written brief, in one afternoon. No developers, no six-month IT project.

    Guardrail applied100% simulated data — nothing real ever entered the tool
    Step 4 · Document & prepare
    The presenter's run-of-show

    The facilitator's cheat sheet — timings, click path, talking points, Q&A answers — was AI-drafted and human-reviewed before you walked in. The draft–review–send habit, applied to ourselves.

    Guardrail appliedHuman review on everything audience-facing — AI drafts, people decide

    Your hands-on exercise starts now

    Pick one task from your own week — a report, a guest note, a schedule — and rebuild it with Copilot using the red-amber-green check. We'll reconvene on the follow-up Zoom session in two weeks to compare results.

    ✓ Noted — the Atlantis University team will bring your department's exercise pack to the follow-up session.