Spaces

Do and don't guide for your practice's AI agent

This guide gives plain do and don't rules for an AI agent answering patient messages, keeping it on admin and handing anything clinical to your team. It is for practice owners writing instructions.

Runs in your browser. Nothing you enter is sent anywhere or saved.

Before you start

Agents follow what you tell them. If your instructions are vague, the agent fills the gaps with its own guesses, and in healthcare a guess can matter. A short list of clear dos and don'ts is the simplest protection you can give your patients and your team.

Read each rule and keep, change or delete it for your own practice. Then paste the final list into your agent's instructions. Review it every few months, and add a new rule whenever a real conversation shows a gap you had not thought of.

Tick each one as you go. 0 of 10 done.

[ ] Do answer questions about hours, address, parking and access
[ ] Do say clearly when a member of staff will reply
[ ] Do keep messages short, warm and in plain words
[ ] Do hand over any message that mentions how the patient feels
[ ] Do point to emergency services if a message sounds urgent
[ ] Don't comment on symptoms, results, diagnoses or medicines
[ ] Don't promise an appointment slot unless the diary confirms it
[ ] Don't share one patient's details with anyone else
[ ] Don't guess at prices, fees or insurance cover
[ ] Don't argue with a patient who is upset, hand over instead

Turning rules into instructions

Write each rule as a short sentence. Keep your wording the same as the list, because plain wording is less likely to be misread. Add one or two real examples, such as a patient asking about a rash, and say exactly what the agent should do.

Test the rules by sending your agent awkward messages. If a reply surprises you, tighten the rule, then test again.

  • Short sentences
  • One real example per tricky rule
  • Test, adjust, test again

FAQ

How many rules should we start with?

Ten clear rules is a good start. Too many rules become hard to follow, so keep the ones that protect patients and add more only when real conversations show they are needed.

Can the agent ever say "that sounds fine"?

No. Even a casual comfort about a symptom can be wrong. The agent should thank the patient, say a member of the team will reply, and let a clinician judge.

Do these rules apply to dental and therapy practices too?

Yes, with small changes. Replace examples with your own, such as tooth pain or a session cancellation, and keep the core rule that clinical matters go to a trained person.

More free tools for healthcare practices

Read the guide: Answer patients' admin questions with an AI agent.