Skip to content

OTA Inbox Triage and Reply Drafter

Example prompt: "When a Booking.com, Airbnb, or Vrbo message forwards into our property inbox, classify what the guest is asking, match the reservation to our 'Bookings' Google Sheet, and draft a reply for the host to sign off and paste back into the OTA's own inbox."

The Problem

A small property running on Booking.com, Airbnb, and Vrbo gets its guest messages forwarded through three different notification addresses with three different formats, and the host opens each one, hunts down which reservation it refers to, taps out a reply that has to be pasted back into the OTA's own messaging window, and then forgets to log the special request the guest mentioned in passing. The cot that didn't get added to the booking notes shows up at 3pm with a baby in the doorway. The 'is parking free?' question gets the same hand-typed answer for the fortieth time this year, sometimes after the guest has already arrived.

How GloriaMundo Solves It

We use the Gmail forwarding the OTAs already do and triage from there. An LLM step extracts the platform, the reservation code (each platform formats it differently), the guest, and the verbatim message. A classifier sorts every message into one of six intents — pre-stay question, check-in detail, special request, in-stay issue, post-stay follow-up, or platform notice — and the workflow does the right thing for each: special requests go onto the booking notes so housekeeping picks them up, in-stay issues are mirrored into the Maintenance Log so the triage workflow takes over, and platform notices are logged silently with no guest reply. Every actionable message gets a Gmail draft written in the host's voice for the host to paste back into the OTA's own messaging window — the audit trail stays inside the OTA. Slack gets a one-liner per message with the intent and the draft link, with red-flag prefixes for in-stay issues. Glass Box preview shows the classification, the matched booking, and the draft before anything is logged.

Example Workflow Steps

  1. Trigger (integration): A Gmail message arrives in the property inbox from one of the OTA notification addresses.
  2. Step 1 (llm): Extract platform, reservation code, guest first name, and the verbatim message text from the email body.
  3. Step 2 (llm): Classify the intent into one of the six tiers.
  4. Step 3 (conditional): For 'platform notice', append to 'OTA Platform Notices' and stop.
  5. Step 4 (integration): Look the reservation code up against the 'Bookings' tab's 'ota_reservation_code' column; mark 'no booking match' if no row found.
  6. Step 5 (llm): Draft a Gmail in the host's voice tuned to the classified intent; save as a Gmail draft.
  7. Step 6 (conditional + integration): For 'special request', append to booking notes; for 'in-stay issue', mirror into the 'Maintenance Log' tab; for 'post-stay follow-up', append to 'Post-Stay Tickets'.
  8. Step 7 (integration): Append to the 'OTA Inbox Log' tab with received_at, platform, reservation_code, matched_booking_reference (or 'no match'), intent, draft link, and a sanitised one-line summary.
  9. Step 8 (integration): Post a one-liner in #front-desk on Slack; red-flag prefix and @owner tag for 'in-stay issue'.

Integrations Used

  • Gmail — the inbound trigger for OTA-forwarded messages and the per-message draft saved for the host to paste back into the platform
  • Google Sheets — the 'Bookings' tab for the reservation match, the 'OTA Inbox Log' for the audit trail, and the mirror tabs (Maintenance Log, Post-Stay Tickets, Platform Notices)
  • Slack — the #front-desk channel for the per-message one-liner with the draft link

Who This Is For

Hosts at properties running on two or more OTAs alongside direct bookings, where the guest messages arrive in the property inbox via the OTA's forwarding and the host currently triages them by reading every email, finding the matching reservation by hand, and pasting a reply back into the OTA's own dashboard. The workflow doesn't bypass the OTA's messaging — it speeds up the read-classify-draft-log cycle so the host can sign off in seconds rather than minutes.

Time & Cost Saved

A property doing forty messages a week across three OTAs spends two to three minutes per message reading, matching, and drafting — that's eighty to a hundred and twenty minutes a week, four to six hours a month. The workflow doesn't remove the host's sign-off — it removes the reading, the matching, the looking up, and the drafting. The bigger gain is the special requests that currently fall through: the cot that wasn't on the notes, the allergy that wasn't on the kitchen list, the anniversary that nobody mentioned to the housekeeper. Mirroring those into the booking notes the moment they arrive is the difference between a guest who feels remembered and a guest who repeats themselves at check-in.