You have 400 reviews telling you exactly what to fix. Nobody on your team has time to read them all. An agent does it in one pass.
Every hotel sits on a goldmine of feedback and acts on almost none of it. Reviews get answered one by one, then forgotten. The pattern across all of them, the thing that would actually move your score, stays invisible.
We build an analysis agent that reads the whole corpus and tells you what to fix.
The workflow has two agents.
Agent 1, the Tagger: reads every review from the last 12 months and tags each by theme. Cleanliness, check in speed, breakfast, noise, value, staff. It also tracks the star rating attached to each theme.
Agent 2, the Analyst: finds the patterns and ranks them by impact. The prompt, verbatim:
From the tagged reviews, identify the three issues that appear most often in reviews rated 3 stars or below. For each, quote two representative guest lines and estimate the rating lift if it were resolved. Rank by how many guests it affects. No generic advice.
The output is not a word cloud. It is a ranked list: the three fixable problems costing you the most stars, in your guests' own words, with the highest impact one first.
One agent turns a year of scattered complaints into next month's operational priority. That is the difference between collecting feedback and using it.
Pull your last 12 months of reviews into one file this week. That export is all the agent needs to start.