AI Analyst
Every review read automatically for sentiment, intent, urgency and topics, each backed by a quote.
What it does
Ratings tell you how customers feel. The AI Analyst tells you why.
Every review that arrives with written feedback is read automatically and labelled with:
- sentiment, meaning how the writing feels: negative, neutral or positive;
- intent, meaning what the customer wanted: a complaint, a question, a suggestion or praise;
- urgency, meaning whether this needs answering today, this week, or whenever;
- topics, meaning what they actually talked about in their own vocabulary, such as wait times, parking, a specific dish or a member of staff;
- a one-line summary of the review;
- Business Health Radar scores, once you have set the radar up.
The result appears on the review card in the inbox, drives the AI filters, and feeds the topic charts in Reports.
Every claim carries a quote
Each topic and each radar score is stored with the exact snippet of the review it came from, and that snippet is verified against the review's own text before it is saved. If a claim cannot be traced back to something the customer actually wrote, it is dropped rather than shown.
This is what turns "customers keep mentioning wait times" into something you can act on: you can click through and read the sentences.
Reviews without text
A review with stars and no words still gets a reading, but it is derived from the rating alone and is marked as such. Star-derived readings never mix into the sentiment and topic statistics, since a sentiment chart that quietly included them would just be a second rating chart.
What it costs you
Nothing. Background analysis is funded by ReviewHive rather than by your monthly token allowance. Your tokens are spent only on the replies you generate. Your plan does set a ceiling on how many reviews may be analyzed. See AI usage and limits.
Turning it off, and erasing what it found
In AI Studio, Owners and Admins can do two things.
Turn the AI Analyst off. New reviews stop being analyzed. Everything already found is kept, so turning it off does not delete anything.
Erase derived intelligence. This permanently removes every analysis from every review in the organization, so the panels in the inbox, the AI filters and the topic charts all go empty. Your reviews and replies are untouched. It cannot be undone, though you can turn the analyst back on and let it re-analyze.
Both controls need the Admin or Owner role.
Accuracy
The analyst is a language model, and language models are occasionally wrong. Two things limit the damage. Every claim is quoted, so you can check it in a second. And the labels come from a fixed set rather than free text, so the analyst cannot invent a category. If a reading looks wrong, tell us. The prompts are evaluated against a graded set of reviews before any change ships.