01 the business
Boutique hotel group
4 hotels, 160 staff
Reviews arrive daily across Google, Tripadvisor and booking sites. Managers know replies matter but they slip to the bottom of the list. Good reviews go unthanked, and the occasional angry one sits unanswered for weeks.
02 before
The problem
Owners know they should reply to reviews, but it falls to the bottom of the list. Good reviews go unthanked and bad ones sit unanswered where every future customer can read them. When replies do go out, they are often rushed, defensive or copy-pasted.
- Most reviews never get a reply.
- Replies that do go out are generic or defensive.
- Serious complaints are noticed late.
03 the build
What the AI does
- Step 1: Collects new reviews from Google Business Profile, Tripadvisor, Trustpilot and Facebook as they are posted.
- Step 2: Reads what the reviewer actually mentioned, such as a staff member, a dish or a late arrival, so each reply is specific.
- Step 3: Drafts a short reply in your voice that thanks, apologizes or explains as needed, without sharing private details.
- Step 4: Flags reviews that mention safety, health, legal threats or a named employee for the owner's personal attention.
- Step 5: Sends drafts to the owner in Slack or email to edit and approve, then posts the approved reply.
paused: waiting for a person to approve
How it was put together
- Collected new reviews from each platform as they were posted.
- Drafted short replies that mention what the guest actually said.
- Flagged reviews about safety, health, legal threats or named staff for the manager personally.
- Sent drafts to managers in Slack to edit and approve, then posted the approved reply.
04 human in the loop
Where a person stays in charge
The owner or manager approves every reply before it is posted. Negative and flagged reviews always need a person to read and edit the draft first.
- Every reply is approved by a manager before posting.
- Replies never share booking details or other private information.
- The tool never writes or solicits reviews, in line with rules against fake reviews.
05 rollout
From kickoff to live in 2–3 weeks
- Week 1Connect review platforms and agree the voice and escalation rules.
- Week 2Draft replies for the last month's reviews and calibrate tone.
- Week 3Run daily for all properties.
06 after
What changes
Every review gets a thoughtful, timely reply without eating into the owner's evening. Unhappy customers see a calm response, and patterns in complaints become easier to spot.
Assume 300 reviews a month across four hotels, at 6 minutes to write a reply and 1 minute to approve a draft.
300 × 6 minutes = 30 hours a month to reply to everything by hand. 300 × 1 minute = 5 hours to approve drafts.
What to measure
- Share of reviews with a reply.
- Median time to reply.
- Manager time spent on reviews.
- Average rating trend.
07 evidence
What the research says
The FTC's 2024 rule bans fake reviews, including AI-generated ones, undisclosed reviews by company insiders and threats used to suppress negative reviews.
US Federal Trade Commission, Final rule banning fake reviews and testimonials (opens in a new tab) · 2024
In a study of 453 professionals doing writing tasks from their own jobs, access to ChatGPT cut the time taken by 40% and raised quality, as rated by evaluators, by 18%.
08 lessons
Mistakes to avoid
- Copy-paste replies. Guests notice and so do future guests.
- Arguing in public. Take disputes offline.
- Generating reviews. Fake and AI-written reviews are banned under the FTC's 2024 rule.
09 questions
Common questions
- Can it reply automatically to five-star reviews?
- It can, if you choose. Most owners still approve every reply because it takes seconds.
- Does it work for clinics?
- Yes, with extra care: replies must never confirm that someone is a patient.