The first principle is restraint. An assistant that tries to do everything does nothing well. Decide on its primary job — usually answering common questions and helping people book — and design around that.

Tip
Tip: Write the assistant's job in a single sentence. If you can't, it's trying to do too much.

Write in your real voice

Patients can tell the difference between a script and a voice. Your assistant should sound like your practice on a good day: warm, plain-spoken, and unhurried. Feed it the words you actually use and the tone you actually keep.

Note
Note: Read the assistant's replies out loud. If they don't sound like something you'd say to a patient at the desk, rewrite them.

Design for the questions people really ask

Map the handful of questions you answer over and over — hours, location, what to expect, insurance, how to book — and make sure the assistant handles those flawlessly before you reach for anything fancier.

Always leave a door to a human

The most important design decision is the handoff. The assistant must recognize when it's out of its depth or when a person is upset, and pass the conversation to a real human gracefully.

Warning
Warning: An assistant that traps people in a loop with no way to reach a person will cost you patients. The escape hatch is not optional.

Set boundaries on purpose

Decide in advance what the assistant should not do — give clinical advice, make promises about outcomes, or discuss anything sensitive. Encode those boundaries clearly.

Navigate to:
ConversationsAI AgentsSettings

Improve it from real conversations

After launch, read the transcripts. The questions it fumbled are your roadmap. A few minutes a week reviewing real conversations turns an average assistant into one patients genuinely appreciate.

Next step

Define the assistant's one job and its handoff rule first, then load it with your most common questions in your own words. When you'd like help training it on your practice, request a guided trial and we'll set it up together.