What the case measured
Replify describes automation at Seattle Sun Tan and Desert Sun, two SST Group brands operating in more than 60 locations in the US Pacific Northwest. The case discusses calls across about 50 salons. The company wanted to reduce repetitive inquiries and handle questions after hours, when staff at the locations were busy with customers.
Replify reports that the system automatically resolved 85.5% of questions about opening hours and 61.9% of questions about treatments. Across those two categories, the figure was 74.4%. This is not a measure of all incoming calls and was not independently audited: the results were published by the vendor. They should not be assumed for another salon without testing.
What a small business can reuse
The best first topics are questions whose answers already exist in current schedules, price lists, and service catalogs: opening hours, basic visit conditions, treatment descriptions, or booking rules. The system should hand off medical restrictions, complaints, payment disputes, unusual requests, and cases with missing information to a person.
The useful metric is not “the percentage of conversations with AI,” but the share of inquiries that reached a genuinely useful outcome. Keep the original question, response, booking or handoff event, escalation reason, and result of subsequent review. Do not count a call as resolved if the model answered but the customer had to call back.
How to run a verifiable pilot
- Select two or three frequent, low-risk topics with a clear source of current information.
- Take one month of inquiries and label each by topic, correct answer, and whether staff were needed.
- Enable a limited scenario at one location or phone line, with a simple handoff to an operator.
- Measure correctness, completed-task share, repeat contacts, complaints, response time, and cost per resolved inquiry.
Schedules, services, and prices need an owner and an update deadline. If a price changes, the assistant must not keep answering from an old version. Booking requires separate rules for confirming a time and checking availability, not merely a well-written response.
Economics and limitations
The vendor links automation to after-hours availability and freeing staff capacity. The case does not provide an independent measure of those effects. Calculate economics per inquiry correctly resolved without a repeat contact, including telephony, knowledge-base setup, quality monitoring, and manual exception handling.
The 74.4% figure is useful as a hypothesis that some repetitive questions can be deflected. It does not prove that AI can take 74% of all calls at another chain. A safe next step is one location, two inquiry types, and a defined escalation path, then compare with manual handling.
