In this article
The pitch has become familiar enough to finish yourself. Every call answered, day or night, in any language, booked straight into the practice management system, intake forms sent before the visit. It is a good pitch and the products largely do what they say. The trouble is that it describes one stage of a journey with five, and it is the stage that was already closest to working.
The front desk got automated first#
There is now a genuine category here. In dental, Viva AI, Hello Patient, Rondah and Adit all sell an AI receptionist that answers calls around the clock and books into the PMS. In aesthetics, Clara AI, My AI Front Desk, Emitrr and Synthflow sell the same shape, often extended to texts and Instagram messages.
The incumbents are buying their way in rather than waiting. Weave, which most practices know as a phone system, acquired TrueLark in May 2025 for a reported $35 million and turned its text and chat automation into Weave AI, and now sells an AI receptionist alongside the phones. When the established patient-communication vendor pays eight figures to add this, the category is not a fad.
This is not a bubble and it is not vapourware. Missed calls are a real and measurable loss, a phone that rings out at seven in the evening is a patient who books somewhere else, and software that answers reliably at three in the morning is genuinely better than a voicemail box. If you have not looked at this category, look at it.
It got automated first for a plain reason. The phone is the most visible failure in a practice. Everyone can hear it ringing, everyone knows it went unanswered, and the fix is legible to a buyer in one sentence. The stages further down the journey fail silently, which is a much harder thing to sell against.
What it looks like from the owner's chair#
Ask an owner where their patients live and the answer is rarely one place. The clinical record is in the PMS or the EMR. The schedule is in the booking system. Text threads with patients are in the phone system. Instagram DMs are on somebody's personal login. Google and Meta lead forms drop into an inbox that is checked when there is a gap. The recall list is a report somebody runs, exports, and means to work.
Adding an AI receptionist to that does something specific and worthwhile: it closes the phone gap. It does not consolidate the rest. You now have a very good answer to the ringing phone and the same six places where a patient can quietly fall out of view.
We are busier than we have ever been, and I could not tell you what happened to the forty people who enquired last month.
That feeling is not a staffing failure. It is what happens when capture is instrumented and everything after it is not.
The five stages after the booking#
A patient is worth something only if they move through a sequence, and the booking is the first step, not the finish line. Every stage below leaks, and each one leaks for a different reason.
Booked to showed
The appointment exists and the patient does not arrive. A reminder helps and is mostly solved. What is not solved is the gap that opens when someone cancels, because a cancelled slot is both an empty chair and a patient with no next step, and the second one is usually forgotten.
Showed to treated
In dental this is case acceptance: treatment is diagnosed, the patient says they will think about it, and nothing follows. In aesthetics it is the consult that does not convert. This is the single most valuable list in either business and the least worked, because working it takes a different message per patient.
Treated to reviewed
The one moment a patient is most willing to say something public about you is the week after a good outcome, and almost nobody asks then. Asking later gets a worse review from a colder memory, if it gets one at all.
Reviewed to referred
A happy patient who has already written a review is the warmest introduction available to any practice, and the ask almost never happens because there is no moment in the workflow where it would.
Referred to returned
The interval passes, nothing notices absence, and a patient who was perfectly happy simply stops being a patient. This is recall, and it is the stage everyone knows about and the one most often run as an annual report rather than a system.
Notice what these have in common. None of them is a phone problem. None of them is fixed by answering faster. All of them are cases where the right person needed a specific message at a specific moment and nobody had the time or the trigger to send it.
Why the wave stops at booked#
Products measure what they were built to measure, and the AI receptionist category grew up around the call. The numbers in every deck are calls answered, after-hours coverage, and appointments booked. Those are the right numbers for the problem being solved. They are also where the reporting ends.
To be fair to the category, some vendors are pushing past it. Hello Patient describes working the recall list. Rondah describes reactivating dormant patients. That is the correct direction and it is worth crediting rather than pretending the whole field stops at hello.
A growth layer around the PMS, not instead of it#
The instinct when six systems hold your patients is to consolidate into one. That instinct is right about marketing and wrong about the clinical record. Your PMS, EMR or POS holds the chart, the schedule and the money, and moving any of those to improve your follow-up is a bad trade at any price.
What can move is everything that is not the record. The enquiry that has not booked. The consult that has not converted. The treatment that was diagnosed and never scheduled. The patient who has not been back in fourteen months. None of that is clinical data, all of it is currently spread across an inbox, a phone, a DM folder and a report, and it is the entire difference between a busy practice and a growing one.
That is the layer Wysera runs. Embeddable forms with qualification rules catch the enquiry wherever it arrives. Six pipeline detectors fire on behaviour rather than on dates, so a patient who has drifted past their interval and a patient with unscheduled treatment are two different signals rather than one undifferentiated list. Those land in a daily queue capped at fifteen items, because a list of four hundred is not a to-do list, it is a reason to close the tab.
The agent drafts each message on a tick schedule with a per-iteration audit trail, and you approve before anything reaches a patient. When somebody is worth a call rather than a message, they queue in the dialer with one-tap disposition so the outcome gets recorded instead of remembered. The chart never moves and the schedule stays where it is.
One week, in order#
Abstractions are easy to agree with and hard to act on, so here is what a week actually looks like in a practice running both an AI front desk and a growth layer behind it.
Monday. Forty-one enquiries arrived over the weekend across the phone, the website form and Instagram. The AI receptionist answered eleven calls and booked six of them. The other thirty enquiries never rang; they were forms and messages. By nine fifteen the queue has drafted replies to all thirty, sorted by treatment interest, and the practice manager approves them from her phone before the first patient sits down.
Tuesday.Two of Monday's consults did not book on the day. They are not lost and they are not in a generic nurture sequence either; they are in a short, specific follow-up about the thing they actually asked about, going out tomorrow.
Wednesday. The unscheduled treatment list produces nine names, each with a different reason for not scheduling. Three get a message about cost and finance options, four get a message about appointment length because they said they were short of time, two get queued for a call because a conversation will do more than a text.
Thursday. Four patients treated last week are asked for a review while the outcome is still fresh, not in a monthly batch six weeks later when the memory has faded.
Friday. The lapsed list runs, segmented by behaviour. Patients who replied to the last message and then went quiet get one thing. Patients who have ignored four get something different, or nothing at all, because a system that never stops is how a practice teaches people to ignore it.
Nothing in that week is clever. Every item on it is work the practice already intended to do and did not have the hours for, which is the only honest description of what this category is for.
So is an AI receptionist enough?#
No, and the reason is not that the products are weak. It is that they are solving the stage that was closest to solved, because it is the stage everyone could hear failing.
There is a version of this where the answer is yes, and it is worth stating plainly. If your phone genuinely rings out, if after-hours enquiries go to voicemail, if the front desk is drowning, then an AI receptionist is the highest-value thing you can buy this quarter and you should buy one before you read another word about growth layers. Capture that is broken is not fixed by better follow-up.
The moment it stops being enough is the moment capture is working and growth still is not. That is a common and confusing place to be, because every dashboard you own is green. Calls answered, appointments booked, chairs full. Meanwhile the treatment plan from March is still unscheduled, the patient from last summer has not been asked back, and nobody has requested a review since the spring.
If every number you report on ends at the booking, then every number you report on describes a quarter of the business.
Spend an afternoon counting three things: enquiries last month that got a reply within the hour, patients with diagnosed treatment who never scheduled, and patients who have not been back in over a year. Whichever number is worst is where your growth is leaking, and it is usually not the phone. There is more on the enquiry side in responding to leads faster and on the retention side in the patient recall playbook.
Frequently asked
Is an AI receptionist enough on its own?
No. An AI receptionist solves capture: answering the call, qualifying the caller, and putting an appointment in the practice management system. That is one stage of five. A booked patient still has to show up, accept the treatment that was diagnosed, leave a review, refer someone, and come back at the right interval. Nothing about answering the phone faster makes any of those happen, and in most practices nothing at all is watching them.
What is an AI front desk?
Software that answers inbound calls, and increasingly texts and social messages, without a person picking up. The current generation qualifies the caller, checks availability, books directly into the practice management system, and can send intake forms before the visit. Vendors in dental include Viva AI, Hello Patient, Rondah and Adit; in aesthetics, Clara AI, My AI Front Desk, Emitrr and Synthflow among others. The category is real and the products work.
What is patient recall automation?
Software that contacts patients who are due or overdue for an appointment without somebody remembering to do it. The useful kind segments by what the patient actually did, so a patient with diagnosed treatment they never scheduled gets a different message from one who simply drifted past a hygiene interval, and it stops chasing people who have clearly stopped responding rather than sending forever.
What does speed to lead mean for a med spa?
How long it takes to give a real answer to someone who enquired, usually through an Instagram form, a Google lead form or website chat. The person asking about a treatment on a Tuesday night is asking two or three clinics, and the first useful reply tends to win the consult. Speed to lead measures the reply, not the ring, which is why a phone system that answers instantly can still sit next to a form enquiry nobody touched for a day.
Does a growth layer replace the practice management system?
No, and one that tries to should be treated with suspicion. The PMS or POS holds the clinical record, the schedule and the money. A growth layer works with enquiry and contact data, which is who asked, what about, what was sent, and what happened next. Keeping the clinical record in one place is both safer and the reason the growth layer needs no migration to start.
How is this different from what the front desk already does?
It is not different in kind, it is different in whether it happens at all. Every practice intends to work the unscheduled treatment list and call the patients who have not been back in a year. Almost none do it consistently, because the waiting room is in front of you and the list is not. The work is well understood and chronically unstaffed, which is exactly the shape of problem software is good at.
Do AI receptionist vendors do recall too?
Some are moving that way and it is worth saying so. Hello Patient describes working the recall list, and Rondah describes reactivating dormant patients. The honest distinction is emphasis and measurement: the category grew up around the call, and what vendors report on is calls answered and appointments booked. Ask any vendor what they measure after the booking, and the answer tells you where the product actually is.
What should a practice fix first?
Whichever stage is leaking most, which is usually knowable in an afternoon. Count last month's enquiries and how many got a reply within the hour. Count the patients with diagnosed treatment who never scheduled. Count how many patients have not been back in more than a year. Whichever number is worst is the first thing to fix, and it is often not the phone.
Try it on autopilot
Keep the front desk. Fix what happens after the booking.
Wysera runs the growth layer around your PMS or POS: enquiry follow-up, consult conversion, unscheduled treatment, reviews and recall, all drafted for your approval. Nothing migrates and the chart never moves.
Running a dental practice instead? See the dental version.