Review365
An AI review management service for restaurants in Australia. I run its sales and client relationships and build its monthly marketing reports.
Review365 turns happy customers into Google reviews: scan a QR code, tap what you liked, AI drafts the review, the customer posts it. My part is the business end of the service: I run the sales and client relationships, and I build the monthly marketing report every subscriber gets, drafted by Claude on my procedures with a human final pass.
The product
Review365 is an AI review management service for hospitality businesses in Australia, and the recording at the top of this page is the product itself: a customer scans a QR code at the counter, taps a few multiple-choice questions about their visit, AI turns those answers into a natural review draft, and the customer posts it to Google in well under a minute. Reviews build up, the profile climbs the map rankings, and unhappy feedback goes to the owner instead of going public.
Won by referral, run in person
Clients come mostly through referrals and word of mouth in Sydney's hospitality scene, not through advertising. I run the sales conversations and the client relationships myself, and once a restaurant signs I operate the account end to end, from the first meeting to each month's report. Ten businesses have signed, four are ongoing monthly engagements, and owning both the winning and the delivering is the business side of this portfolio at its most literal.
The monthly loop
Every month the same loop runs: keyword rankings tracked against local competitors, Google Business Profile insights compared over stable windows so the numbers stay honest, and an AI visibility check covering how the business shows up in LLM answers. Gaps and wins land in a report with charts and recommendations the owner can execute.
LLMs in production, with a human final pass
Generation runs on Japanese SOPs executed by Claude: a thirteen-step procedure with preflight checks and schema validation drafts each report, and I make the final call before anything reaches a paying client. Operating LLM output at client quality, month after month, is the discipline this service taught me.
A real report, as delivered
The pages in this case study are from an actual monthly report as delivered to a paying client (identifying details masked). The KPI header the owner reads first, the month's summary in plain language, keyword movement against local competitors, and a prioritised list of moves. When the numbers dip, the report says so plainly; an honest bad month is what keeps the good months credible.
