Almost every article about AI receptionists is written by a company selling one. This is the honest version, from a studio that builds them for clinics, including what they shouldn’t do and where Meta’s new Muse fits in.
The short answer
An AI receptionist is excellent at answering every call, booking appointments and answering routine questions, day and night. It should never give medical advice, triage or handle emergencies, and it should always say it’s an AI and hand off to a person quickly. Done well, it takes phone pressure off your team. Meta’s Muse is a different kind of tool: it helps you with admin and marketing, but it doesn’t answer patients.
What it can and can’t do
- Answer every call, including after hours and when the desk is busy
- Book, move and cancel appointments in your booking system
- Answer routine questions: hours, prices, insurance, parking, what to bring
- Take a clear message and alert your team right away
- Send a confirmation text after the call
- Handle several calls at once, so nobody hears a busy signal
- Give medical advice or tell a patient what’s wrong
- Handle emergencies. It should tell callers to hang up and call 911
- Decide who is urgent. Triage belongs to a trained person
- Pretend to be human
- Handle upset or complex callers without a fast human hand-off
- Replace the warmth of your front desk for patients in the clinic
What a good setup looks like
Here’s the call flow we built for a busy chiropractic clinic that kept missing calls. It books straight into Jane App.
Read the full case study. The two details that matter most: callers are told they’re talking to an AI, and a human is never far away.
What about Meta’s Muse?
Meta launched Muse for Small Business on September 29, 2026, in the US and Canada (TechCrunch). Many clinic owners are wondering if it can answer their patients. Here’s what it is, and isn’t.
It connects to tools like Instagram, Facebook, QuickBooks, Canva and Stripe to help you plan, analyse and draft: social posts, ad ideas, emails. It’s free with usage limits, with paid tiers for heavier use.
It doesn’t answer customers or phone calls. Meta’s separate tool for customer messages, Business Agent, reportedly excludes health businesses (eesel review).
According to the same review, Muse uses your data for AI training by default (you can turn it off) and runs on consumer-style terms without a business data agreement. Use it for marketing and admin, and keep patient information out of it.
What patients want from the phone
Patients aren’t against AI, but they want a person when it matters. In a 2026 survey of more than 2,000 US adults, nearly a quarter of people who reached AI on a health call gave up without getting the help they needed (Medical Economics). For test results and emergencies, a clear majority wanted a human.
- Say it’s an AI in the first sentence. Honesty builds trust; pretending backfires.
- Offer a person at any time, and make the hand-off fast: a call-back within minutes, not days.
- Keep it to logistics: booking, hours, prices, directions.
- Route anything urgent immediately: “If this is an emergency, please hang up and call 911.”
- Listen to real calls in the first weeks and fix what confuses patients.
Privacy rules in Canada
An AI receptionist hears personal information, sometimes health information. Two sets of rules apply.
The federal Privacy Commissioner says organizations must tell callers they’re being recorded, “clearly state the purpose of the recording and ask for their consent,” and offer an alternative if they object.
Privacy Commissioner of CanadaOntario’s privacy commissioner released AI guidance for health custodians in early 2026: complete a privacy impact assessment first, keep trained humans reviewing outputs, and expect vendors to report problems promptly.
Baker McKenzie summaryHi, you’ve reached [clinic]. I’m the clinic’s AI assistant, and this call is recorded so our team can follow up accurately. I can book or change appointments and answer general questions. To speak with a person at any time, just say ‘team member’. If this is an emergency, please hang up and call 911.
10 questions to ask any AI receptionist vendor
Where is call data stored, and for how long?
Ask for the country and the retention period in writing.
Is our data used to train your AI?
The answer should be no, or you should be able to switch it off.
Will you sign a data agreement with us?
Including what happens to data if you cancel.
How does it connect to our booking system?
Direct integration, or does staff re-type bookings?
How fast does a human get the message?
Text, email or alert? Within minutes?
Can callers reach a person at any time?
And what happens after hours?
How does it handle emergencies?
It should direct callers to 911, never assess them.
Can we hear real call recordings?
From a clinic like yours, with permission.
What does it cost, all in?
Setup, monthly fee, per-minute charges, phone number.
Who fixes it when something goes wrong?
A person you can reach, not just a help centre.
Is an AI receptionist right for your clinic?
Questions clinic owners ask
Will patients be upset to talk to an AI?
Some will, especially for stressful calls. That’s why the AI should say it’s an AI from the start, keep calls about logistics, and pass anything sensitive to a person quickly.
Can an AI receptionist book in Jane App?
It can, with the right setup. We built one for a chiropractic clinic that books straight into Jane through backend automations.
Is it allowed under Ontario privacy law?
It can be, if you set it up carefully: tell callers about recording and why, collect only what’s needed, know where data is stored, and have a proper agreement with the vendor. Health information custodians should complete a privacy impact assessment first.
Can I use Meta’s Muse as my receptionist?
No. Muse helps the owner with drafting and planning; it doesn’t answer patients or phone calls. Keep patient information out of it.
Will it replace my front desk?
It shouldn’t. The best setups take the phone pressure off your team so they can focus on the patients in front of them.
How much does it cost?
It depends on call volume, integrations and setup. Compare it with what missed calls cost you: if even a few new patients a month were reaching voicemail, the maths usually works.