BookingFlow AI
BookingFlow applies AI in focused, useful places rather than as a gimmick. This page maps where it appears, what it does in each spot, and the principles behind it.
Where AI appears
| Feature | What the AI does |
|---|---|
| Guest inbox | Drafts replies with booking facts + your tone; summarises unread mail into to-dos; pulls live weather, events and traffic context for the guest's dates |
| Reports | Answers plain-English questions about bookings, revenue, occupancy and meter data, with charts |
| Reviews | Rates sentiment per review; one-line summaries; strengths, weaknesses and highest-impact focus areas |
| AI Agents | Runs processes you describe — reads freely, writes only with your approval |
| Reconciliation | Classifies Amazon orders to Xero account codes (conservatively — unknowns default to General Expenses) |
| Voice assistant | Understands questions and creates bookings by conversation |
How the AI knows your park
The AI features read the data you already have — bookings, allocation, meter readings, reviews — and the Communications Preferences you set (site description, tone of voice, sign-off). No separate setup.
Principles
- You stay the host. Drafts are suggestions until you press send; reports describe what happened, they don't act; agents' writes need your approval.
- Facts first. Where the AI speaks about a booking, it's fed the real booking — guest, accommodation, dates, party, price — so it doesn't invent details.
- Context where it helps. For guest replies, the AI fetches real weather, local events and traffic for the guest's dates — and shows you the context it used.
- Conservative with money. Anything touching accounts (invoices, Xero) arrives as a draft for human approval, and AI classification of unknown items defaults to safe accounts.
Tips for better results
- Fill in the Communications Preferences — a good site description makes every drafted reply better.
- Ask Reports specific questions — "revenue by month for 2026" beats "how are we doing" when you want comparable numbers.
- Give agents narrow instructions — one process per agent, with only the actions it needs.
- Re-analyze Reviews after syncing — insights refresh when you ask, not automatically.