AI Coworker Case Study · Booking Steward
Stage 3 · EvaluationStage 4 · Booking2 properties, Upstate New York

"Is the pool heated?" got an auto-reply and nothing else. Since the AI coworker started reading every inquiry, no question has been left that way.

The property's PMS sends an instant, polite auto-reply to every inquiry. It looks like an answer in the inbox, but it never answers the guest's actual question. Three real questions slipped through that way in August. None have since an AI booking steward started checking every conversation.

67
Inquiries reviewed, Jul 8 to Sep 25
3
Real questions that got only the auto-reply before Sep 12
0
Left on the auto-reply since
100%
Of replies approved by a person before sending

Before: the auto-reply hid the gap

Each of these guests got the PMS's automatic "thanks for your interest" message within seconds. None got a real answer from anyone afterward. Questions are paraphrased and names removed.

InquiryWhat the guest askedReal answer ever sent?
Aug 18, morningIs late checkout an option?No, only the auto-reply
Aug 18, eveningIs the pool heated? What are check-in and checkout times?No, only the auto-reply
Aug 19, afternoonWhat time can I get in?No, only the auto-reply

Each of these is an easy question from someone deciding whether to book. On Airbnb, host responsiveness is one of the things search ordering considers, and Superhosts must respond to 90% of new messages within 24 hours (article 39, article 829). An auto-reply counts as a response in the inbox, which is exactly why nobody noticed.

What the AI coworker does

  1. Checks the PMS every 30 to 60 minutes for any conversation where the guest spoke last and no real host reply followed. Auto-replies don't count.
  2. Opens a queue item for each one and marks it red, amber or green by urgency.
  3. Drafts a researched reply from the property's knowledge base, answering the question asked and the next one or two the guest is likely to have.
  4. Posts the draft in the team's Slack, labeled NOT SENT. It sends only after a person approves it with one tap.

After: nothing dropped

From September 12, when the booking steward started, to September 25, eight inquiries got real replies, and none were left with only the auto-reply. One more arrived on the evening we pulled the data and was already in the approval queue.

What we can and can't claim

  • This is about nothing being dropped, not about speed. The median time to a real reply was about 6 hours before and about 12 hours after, because every reply now waits for a person to approve it. We are not claiming faster replies.
  • We are not claiming lost bookings. The PMS can't tell us whether those three guests booked somewhere else. They are three unanswered questions, not three lost reservations.
  • Short window. Thirteen days and eight inquiries after the start date, against nine weeks before it. We will update this page as the sample grows.
  • How we checked. All 67 inquiries and their message threads were pulled from Hospitable on September 25, 2026. We checked the three by hand: each thread holds the guest's question and the automated reply, and nothing else from the host.

Cost

Real numbers from the account's usage records, pulled September 25, 2026, at the account's rate of $2.50 per 1,000 AI credits.

  • A normal week (September 19 to 25): $49.80 (19,918 credits). That is the morning queue brief ($11.62 for the week), the weekly review ($0.27) and every drafting and approval conversation in the guest-messaging channel ($37.91).
  • The inquiry check itself is free. It runs as a plain script every 30 to 60 minutes and uses no AI credits.
  • Roughly $2 per guest message sent, all in. About 25 guest messages went out that week. This over-counts inquiries on purpose: the same channel also handles booked guests, arrival forms and owner questions, and we did not try to split them.
  • The first 13 days, setup included: about $205.

Not counted: the time of the person who approves each reply.

Where Book Them Twice fits

This work ran on the property's PMS with an AI coworker. Book Them Twice was not involved. Inside Book Them Twice, the matching automation is at the Fill Your Nights level: every inquiry captured, a text alert to you, and a 30-minute reply reminder. At the Book Them Twice level, recovery emails go to inquiries that stopped before booking. Book Them Twice makes sure you know about every question; the AI coworker writes the answer.

See the Evaluation stage on the guest journey →

How many of your inquiries got only the auto-reply last month?

Get a demo →

All case studies