Review Automation for a Single Hotel
Overview
This project automates the process of analyzing hotel reviews to identify accessibility-related insights for a single property.
Using AI and workflow automation, the system turns hundreds of guest reviews from TripAdvisor, Booking.com, and Google Reviews into actionable reports — revealing which accessibility features guests appreciate most, and which areas need improvement.
The goal was simple: make accessibility feedback measurable, structured, and report-ready.
Challenge
Able2Global wanted real data to be used as their offer and lead magnet but they needed a tangible and real-world data on how much hotels are losing due to the lack of accessibility functions. So I was brought in to create the solution
Industry
Consulting
Client
Able2Global
Service
Automation
Date
September 2025
Solution
The solution was to create a system that automatically scrapes a given hotel name from 3 different sources - Google reviews, Booking.com, and TripAdvisor. Once collected each review will go through OpenAI (4.1) for categorization and checking if they are accessibility related. Once done a report will be sent to the client containing the top 3 accessibility issues and an AI generated solution for each.
How it works
1. Collects reviews from TripAdvisor, Google, and Booking.com. The architecture uses a polling technique wherein it triggers the Apify Task to scrape thousands of reviews from a specific hotel. Due to it's large data being scraped I had to asynchronously do it which means I had to poll (check) every X minutes if the Apify Task was done. Once all three are done we move to the next phase.
2. Using AI (OpenAI 4.1 Model) to check if a review is accessibility related and whether if it is positive or negative. If you noticed there are two AI nodes since I had to make sure only accessibility reviews were being moved into the next stage. This serves as a "contextual" check and to lessen the hallucination that AI tends to give when handling large amount of data.
All reviews will be saved into the database , in this system I used Supabase. Reviews that are accessibility related it will have its own table within the database.
4. From the database, each review will be categorized and fetched. All reviews that are categorized as X will go through a sub-flow automation (another automation) that gets the main pain points and generates the recommended solutions.
Additional data for the final report will be created such as the total reviews, total positive and negative reviews, and more.
6. Once all data are ready a final report will be created through Google Docs and exported to the client.
