Hosto's AI reads incoming guest messages and generates replies using your property details, conversation history, and booking context. You control how much autonomy it has.
When a guest sends a message, the AI considers:
Property information -- description, amenities, house rules, location, and neighborhood
Conversation history -- what's already been discussed
Message intent -- whether the guest is asking a question, making a request, or raising a concern
Booking context -- reservation dates, guest count, and status
This means it can accurately answer common questions (Wi-Fi password, check-in time, parking directions) without you typing the same thing repeatedly. Better property listings lead to better AI responses.
Mode | What happens |
|---|---|
Controlled | AI suggests replies. You review, edit, and send. |
Autopilot | AI sends replies automatically when confident. Escalates when unsure. |
Escalated | AI pauses. Waiting for you to handle it personally. |
Disabled | No AI. You handle all messages manually. |
Most hosts start here. The AI generates multiple response options that appear as clickable snippets above the reply box.
Click a snippet to load it into the reply field
Edit the text to adjust wording or add a personal touch
Send when you're satisfied
Or ignore the suggestions entirely and write your own reply. See Reading and Replying to Guests for the full workflow.
The AI calculates a confidence score for each response:
80% or higher -- sends automatically (with a brief delay so it feels natural)
60-79% -- escalates with a suggested response for you to review
Below 60% -- escalates for you to handle manually
Auto-sent messages are labeled AI in the thread so you always know what the AI sent on your behalf.
Autopilot works best for high-volume properties, routine questions, and properties with thorough listings that give the AI plenty of information to draw from.
Certain situations always trigger escalation:
Damage reports -- a guest mentions something broken or not working
Payment or refund issues -- questions about charges, billing, or money
Guest complaints -- expressions of dissatisfaction or frustration
Low confidence -- the AI can't generate a response it trusts
When a conversation is escalated, the thread switches to Escalated mode, you get an email notification, and the thread is flagged in your inbox. The AI may still provide a suggested response, but nothing sends automatically.
Handle the situation, then switch the conversation back to Controlled or Autopilot when you're ready. See Setting Up AI Mode for how to change modes.
Tip: Start with Controlled mode on all properties. Watch the AI's suggestions for a few days. Switch to Autopilot on properties where it consistently gets it right.
