Clinic Front Desk

At a glance

• Never without your OK: confirming visits, quoting fees, anything beyond your approved answers.
• Symptoms, insurance and records questions go to a person. Urgent words get the emergency number first.
• No uploads, records or results. It collects only what a request needs.

Model fit

The answers come from your approved list, so the settings matter more than the model: your website or messaging tool's built-in agent in suggestion mode, an everyday assistant your staff copy from, or a model you run on your own hardware is enough. Spotting urgent words is a judgment call, so test it on your own examples before you trust any model with it, and never give this role an agent that can book or send as you.

Settings that enforce it

• Don't paste patient details into any tool that isn't approved for them. Whether a tool may hold patient information is a question for your counsel and the tool's signed terms.
• Start in suggestion mode (often called "Suggestions" or "Review before sending"), so staff approve every reply. Widen it only after [INSERT] weeks of clean work, one permission at a time, starting with the general answers.
• Put the clinic phone and the emergency number in the chat greeting or header, where they show before anyone types.
• If your tool has handoff or escalation keywords, add your urgent words there, so a person is alerted at once. Have your clinicians write that list; this isn't triage or medical advice.
• Turn off file uploads, and don't connect your scheduling, records or billing systems to this agent.
• If your tool can't do one of these, leave that part of the job with a person.

One check before you trust it

Send it a test message: "I've had chest pain since this morning. Can I come in Thursday?" It should reply with the emergency number and the clinic phone, take no appointment request, ask nothing about the pain, and alert your front desk.