The AI Advisor
Ask in plain language and the advisor answers from twenty-one named read-only queries of your own clinic's records — it picks the two or three a question needs, every one is a read (it cannot book, cancel, send or change anything), the question, user, tools and duration are logged, emails and phone numbers are stripped before the model sees a result, Stop aborts the call with nothing saved, and it is included in every plan.
2705
Twenty-one sockets, every one a read
Operations (clinic overview, appointment KPIs, schedule density, operational patterns, cancellation deep-dive, recurring series health), Money (revenue summary, expense overview, insurance claims, unpaid appointments, service menu analysis), People (doctor stats, patient search, patient encounters, medications, demographics) and Growth (review insights, booking source analysis, communication stats, intake form analysis, platform benchmarks). Ask for a tool that isn't on the panel and the dispatcher hands back "Unknown tool." — it reaches no code. Two sockets check their table exists first and say "Benchmarks table not deployed yet." rather than invent a number.
It reads. It never writes.
It can tell you what a refund did to the month; it cannot issue one. Where it surfaces a next step it is a link a person clicks.
What is logged and what is hidden
The question, the requesting user, the tool names and the duration are logged. Emails and phone-shaped digit runs are replaced with [redacted] in every tool result before the model sees it, and patients are referred to by their Qlynic id. Press Stop before the answer is persisted and there is no message saved, no chat row, no usage deducted and no tool-log line.
The trace is not proof
Each answer says which reads built it and how long each took. That is a trace of sources, not a verification of the prose — nothing recomputes a figure the model writes.
Price
Included in every plan; it draws on the clinic's AI units. Powered by Claude.
Where to look