This guide shows you how to train EnsoAI to answer guest questions accurately and consistently, so you can achieve the confidence level of enabling AutoPilot and let EnsoAI handle responses within your set conditions and guardrails.
These tools are part of the EnsoAI V3 (CoPilot) update, now rolling out in waves, starting with users who requested early access.
What “training” means in Enso Connect
EnsoAI learns from structured sources in your account:
Listing data pulled from your PMS (amenities, Wi-Fi, beds/baths, etc.)
Per-listing details stored in Enso (you can open/edit these in the Listings page)
AI-specific sections: Standard Operating Procedures and Policies
Your team’s feedback in the Inbox (accepting/editing suggested replies and adding missing facts)
The more complete and up-to-date your listing and policy content is, the better EnsoAI performs.
The 3 places you’ll “train” EnsoAI
From the Inbox (fastest for day-to-day): This option allows you to seamlessly address any issues during live operations and fill in any gaps without having to leave the conversation. You can also access the purple suggestion bubble in both Inbox and Sandbox. Learn more about this type of training here.
In the Knowledge Base (single-pane view): This is the perfect place to audit coverage, identify any gaps, and gain a better understanding of where information is located. Learn more about the AI Knowledge Base here.
In the Sandbox (bulk “agent” updates & testing): This type of training is currently limited to individual properties, but we are exploring the possibility of implementing a "copy to many listings" feature. Learn about our Sandbox here.
Setting Up: What types of content should you add?
Facts that drive 80% of questions: check-in details, parking, Wi-Fi, beds/rooms layout, appliances (e.g., rice cooker), A/C and heating, pet policy, trash/recycling day, pool/hot tub rules, house rules.
“Always say it this way” content: tone preferences, signatures, disclaimers.
Operational specifics: SOPs for maintenance/lockouts, after-hours support rules, fee logic, upsell rules.
Room configuration (bedrooms/bathrooms) is supported. If answers differ by listing, add them per property.
Improving answers over time (accept vs. edit)
When EnsoAI proposes a Suggested Reply in the Inbox:
Click Reply (without edits) → strong positive signal that the suggestion was correct.
Edit first, then send → the team reviews these to tune accuracy (e.g., signature formatting, phrasing). Edits don’t automatically save as new knowledge yet - use the suggestion bubble if the content itself was wrong or missing.
💡 Why it matters: Your Accept/Edit patterns drive the internal review queue. Once accuracy for a category stays high, we can safely enable AutoPilot for that category.
AutoPilot & categories
Autopilot will roll out per message category once performance thresholds are met (e.g., quick “thank you,” pre-check-in guidance, etc.).
Safer categories (e.g., short acknowledgements) arrive first; sensitive flows (e.g., pre-check-in with access codes, time-bound upsells) require higher confidence.
💡 How you’ll know you’re ready: Rising acceptance rate of suggested replies (fewer edits) in a category. Your CSM/Enso team monitors this and will advise when to flip AutoPilot on.
Recommended rollout plan
Pick 5–10 highest-volume listings.
Audit in Knowledge Base (read-only) to spot obvious gaps: Wi-Fi, rules, parking, room layout, trash day, appliances.
Bulk add in Sandbox (Enso AI Agent) → paste a standardized “facts pack” per listing; confirm destinations.
Verify in Sandbox (Guest Simulation) with your top 20 guest questions.
Use CoPilot in your guest messaging and:
Accept correct Suggested Replies.
Use purple suggestion for missing facts.
Edit phrasing where needed (signature, tone).
Meet with your CSM to review acceptance rates by category and switch AutoPilot on where confidence is high (after it is released to all users).
FAQs and Feedback
Do you have any more questions or feedback about our AI?
Check out our Frequently Asked Questions page here.
You can also contact our support team at [email protected].