Per call, an AI receptionist is far cheaper: a few hundred dollars a month against a full salary. But it only covers the phone. For most dental practices the cheaper answer is AI for overflow and after-hours calls alongside the existing team, not AI instead of a person. A hire still wins where check-in, payments and nervous patients fill the day.
The comparison goes wrong when owners compare an AI subscription with a whole receptionist's salary, as if the job were only answering calls. At the front desk of a busy practice, the phone is often less than half the work. Swap the person for software and the in-person half lands on the dentists, the nurses or the one receptionist who's left. The right comparison is the cost of covering the phone with AI plus the cost of covering everything else, against the cost of a person who does both.
What a dental front desk actually does in a day
Before comparing anything, list the work. Here's an illustrative split for one full-time receptionist at a four-surgery practice; your practice manager can fill in your own in an afternoon by noting tasks in 30-minute blocks for a week.
| Task | Share of the day | Can an AI receptionist do it? |
|---|---|---|
| Answering calls: bookings, moves, questions | 35% | Mostly, if it's connected to the diary |
| Greeting, checking in and checking out patients | 20% | No |
| Taking payments, explaining costs and payment plans | 12% | No for in-person; partly by phone |
| Diary juggling: fitting emergencies, moving long appointments | 10% | Poorly: needs clinical and practical judgement |
| Following up treatment plans and unconfirmed appointments | 8% | Partly, by text and email |
| Forms, medical history updates, consent paperwork | 8% | Partly, with digital forms |
| Complaints, anxious patients, unhappy relatives | 4% | No |
| Chasing lab work, deliveries, other admin | 3% | No |
In this example, the AI can take most of about 35% of the role and part of another 16%. The remaining half or more still needs a person at the desk. That single table usually settles whether "replace" is realistic. The tutorial on which dental front desk tasks AI can take over goes through each task in more depth.
The log itself needs no software. A sheet on the desk with one row per half hour is enough, filled in as the day goes. An illustrative Monday morning extract:
| Time | What filled the half hour | Phone or desk? |
|---|---|---|
| 8:00-8:30 | 11 calls: 4 pain calls wanting today, 5 check-up bookings, 2 cancellations; 3 more rang off unanswered | Phone |
| 8:30-9:00 | Checked in 6 patients, two new-patient forms on paper, 4 calls answered | Both |
| 9:00-9:30 | Fitted two pain patients into the diary after asking the dentist; rang both back | Diary |
| 9:30-10:00 | Payment plan explained at the desk for a crown; 3 calls to voicemail meanwhile | Desk |
Even these four rows tell a story: the lost calls cluster at opening and whenever the receptionist is busy at the desk, and the pain calls needed a dentist's say-so that an AI couldn't have given. A week of rows like these gives you the percentages for your own version of the table above.
Side by side on what decides it
| Criterion | AI receptionist | Front desk hire |
|---|---|---|
| Monthly cost | About $80 to $900 per location, plus review time | A salary plus employer on-costs, recruitment and training |
| Hours covered | 24 hours a day, weekends and lunch | Opening hours, minus breaks, holidays and sickness |
| Monday-morning peak | Answers many calls at once | One call at a time; the rest queue or hang up |
| In-person tasks | None | All of them |
| Emergencies | Follows escalation rules; can't judge beyond them | Judges, and can walk to the surgery and ask |
| Nervous or upset patients | Polite, but flat; some callers hang up | A good receptionist is the practice's reputation |
| Fitting a long appointment into a full diary | Weak unless rules are very detailed | Strong with experience |
| Typical failure | A confident wrong answer, repeated consistently | Slips when rushed or tired; varies by person |
| Time to productive | 2 to 6 weeks of setup and tuning | Weeks to months, depending on experience |
| Data obligations | Call recordings and transcripts of health details held by a vendor | Existing practice systems and training |
Vendor prices for the AI column, from published rate cards and dental platform quotes, are broken down in the companion tutorial on what an AI receptionist costs a dental practice.
Running the numbers for your own practice
Two formulas, filled with your own figures:
Cost of a hire (first year)
= annual salary
+ employer on-costs (payroll taxes, pension, holiday cover)
+ recruitment (advert, agency fee, interview time)
+ training overlap (weeks a colleague spends shadowing)
Cost of AI covering the phone (first year)
= subscription x 12
+ setup fee and integration
+ staff review time (hours per month x hourly cost x 12)
+ callbacks your team still makes
+ cost of covering the in-person work some other wayThe last line of the AI formula is the one that gets left out, and it's often the largest.
A worked decision: replacing a part-time leaver
A five-surgery practice has two full-time receptionists and one part-timer on 25 hours a week, who is leaving. The phones are the sore point: Monday mornings and lunchtimes lose calls, and after-hours calls go to voicemail. All the figures below are illustrative; put your own in.
Option A: replace like for like. Say the part-time salary is $22,000 and on-costs add 20%, so $26,400, plus $1,500 for recruitment and three weeks of training overlap. First year: about $27,900, before counting the colleague time spent training. The phone problem stays exactly as it is, because the practice still has the same number of people answering at peak.
Option B: AI only, no hire. A dental platform at $550 a month including integration ($6,600 a year), $1,000 setup, and six hours a month of review at $30 an hour ($2,160): about $9,800 in year one. But the practice manager's log shows the part-timer spent about 10 of her 25 hours on the phone. The other 15 hours (check-outs during the afternoon rush, payment plans, forms) would fall on the two full-timers, who are already stretched. Realistically they'd need paid overtime or the practice would feel it at the desk.
Option C: a smaller hire plus AI on overflow. A 15-hour-a-week receptionist for the afternoon rush ($13,200 salary, $15,840 with on-costs, plus $1,500 recruitment: about $17,300) plus the AI taking overflow after four rings, lunch and after-hours (about $9,800). First year: about $27,100.
Option C costs almost exactly what A does, and it fixes the phone problem that A leaves in place. Option B is the cheapest on paper and the one most likely to cost staff goodwill. The practice chose C and set a review at 90 days, measuring abandoned calls, bookings made by the AI and overtime hours. That's the typical shape of this decision: the cash saving is small or nil, the service improves, and the real gain is the calls that used to be lost. The tutorial on measuring an AI receptionist's return in 90 days sets up those measurements.
The day-90 readout for that practice might look like this (illustrative):
| Measure | Month before | Month three |
|---|---|---|
| Calls rung off unanswered | 160 | 35 |
| Bookings made by the AI | n/a | 210 |
| New-patient bookings | 38 | 52 |
| Front desk overtime hours | 14 | 3 |
| Complaints mentioning the phone | 4 | 2, both about the AI voice |
The line that justifies the choice is new-patient bookings, and it needs a sanity check before anyone celebrates: was month three a season when new-patient enquiries rise anyway? Compare with the same month last year. The two complaints about the AI voice are worth reading in full, because if both came from older patients, the "press 0 for the team" option belongs in the greeting now, not at the next review.
When the hire wins
- Your problem is at the desk, not on the phone. If patients queue at check-out while the phone is manageable, software doesn't help.
- You have lots of treatment-plan and payment conversations. Explaining a $4,000 implant plan and payment options is sales and reassurance, done best face to face. AI can help prepare the explanation (see explaining treatment plans with AI) but not replace the person having the conversation.
- Your patients skew older or anxious. Some callers will hang up on an AI voice. If your call data shows that, the saving evaporates in lost bookings.
- Your diary is complex. Multiple clinicians, long appointments, sedation sessions and lab-dependent bookings need judgement no set of rules fully captures.
When the AI wins
- Calls are lost at peaks and after hours. A person can't answer three calls at once; an AI can. Every missed new-patient call is a patient who may book elsewhere.
- A large share of calls are routine. Opening hours, parking, "what time is my appointment", check-up bookings. These are the calls AI handles well.
- You can't recruit. If a vacancy has been open for months, AI on overflow keeps service up while you search, whatever you decide long term.
- Your team is spending hours on callbacks. An AI that books directly removes the voicemail-then-callback loop.
The lost-call argument is worth putting into numbers, because it's usually the biggest line and the one least often measured. Your phone system's report will show unanswered calls; listen to a sample of voicemails to estimate how many were new patients. As an illustration: 120 unanswered calls a month, of which one in ten is a new patient who doesn't ring back, is 12 lost new patients a month. If the AI books even half of them and each first visit is a $150 exam and hygiene appointment, that's about $900 a month before any treatment that follows, which more than covers the $550 monthly subscription in the worked decision above. Check the figures against your own fees and call log; the method matters more than these numbers.
The split most practices end up with
Very few practices end up at either extreme. The arrangement that tends to last gives each side the work it's good at, with clear rules for the handover between them.
The AI takes: every call outside opening hours; calls the desk doesn't answer within four rings; the lunch hour; routine bookings and moves for check-ups and hygiene; and the questions with fixed answers, such as prices for a new-patient exam, parking, access and finance options. It sends a text summary of every call it handles to a shared inbox, so nothing lives only in the AI's log.
The team keeps: everything in person; any call mentioning pain, swelling, bleeding, trauma or a lost crown; treatment-plan and payment conversations; complaints; and any caller who asks for a person, immediately and without argument. They also own the morning review: ten minutes reading yesterday's AI bookings and summaries, fixing anything odd before the patient arrives.
A typical morning's catches in the first few weeks, all illustrative: a new-patient exam booked into a hygienist's slot, because both were labelled "30 min" in the diary; a child booked with a dentist who doesn't see children; and the same patient booked twice, once under a date of birth given day-first and once month-first. None of these is dramatic, and each one would have become an awkward conversation at the desk. Each also points to a fix in the diary set-up rather than in the AI: appointment types named by purpose, clinician rules that say who sees children, and date-of-birth confirmation read back to the caller in words ("the 4th of March").
The awkward middle: calls from someone other than the patient. "Can you tell me when my mum's appointment is?" sounds routine, but it asks the AI to share a patient's details with a third party. Set the rule before go-live: the AI takes a message and a callback number, and the team decides what can be shared under your practice's confidentiality policy. The same goes for a partner cancelling on someone's behalf or a school ringing about a pupil.
The handover: when the AI passes a call on, it tells the caller who will ring back and by when, and the team keeps that promise. A callback window that slips from "within the hour" to "tomorrow" is how patients lose faith in the whole arrangement, however good the AI's side is.
What patients notice, when it works, is that someone always answers and that the person at the desk has time to talk to them. That second part is the argument for keeping people on the desk: the AI buys them the minutes.
Two calls, handled both ways
A pair of illustrative calls shows where each side is strong. The first is routine; the AI handles it well.
Caller: Hi, I need to book my six-month check-up.
AI: I can help with that. Can I take your date of birth to find your
record? ... Thanks. I have Tuesday 14th at 9:10 or Thursday 16th
at 4:30 with your usual dentist. Which suits you?
Caller: Thursday.
AI: Booked for Thursday 16th at 4:30. You'll get a text confirmation
and a reminder the day before.The second is where rules run out:
Caller: My crown's come off. It doesn't hurt but I'm flying out on
Saturday for three weeks and I'm worried.
AI: I'm sorry to hear that. As you're not in pain, the next routine
appointment is Wednesday 22nd. Shall I book it?The AI followed its rules: no pain, no swelling, so not an emergency. An experienced receptionist would hear "flying out on Saturday", check with a dentist and squeeze in a ten-minute recement on Friday. The fix is an escalation rule: any caller mentioning travel, an event or a deadline goes to a person, as does any lost crown or filling. After the change:
AI: Thanks for telling me about your trip. I'm going to pass this to
the team so they can see whether we can fit you in before
Saturday. Someone will call you back within the hour, before 5pm.Rules like that come from reading transcripts every week in the first month. The tutorial on call scripts and escalation rules has a full set to adapt.
Signs in the first month that you chose wrong
Track five numbers weekly after any change: abandoned calls, new-patient bookings, callbacks your team makes, patient complaints mentioning the phone, and front desk overtime. If abandoned calls fall but callbacks rise, the AI is taking messages it should be resolving; check its diary access. If complaints rise among older patients, offer "press 0 for the team" in the greeting and keep it. If overtime rises after cutting a role, the in-person half of the job was bigger than your log suggested, and a small hire will cost less than burning out the people you have.
Further reads
- AI or a New Hire? How to Decide Before You Recruit — The general version of this decision for any role you're about to recruit.
- Winning Back Lapsed Dental Patients With AI Messages — Using AI messages to bring lapsed patients back.
- How Dental Practices Use AI for Recalls and Appointment Reminders — Cutting the rebooking calls that fill the front desk's phone time.
- AI Receptionist or Answering Service for a Booking-Based Business? — A third option: a human answering service for overflow.
- Dental Practice AI Mistakes: Consent, Data, and Over-Automation — Over-automation and consent mistakes practices make with AI.
- How to Talk to Staff Who Fear AI Will Take Their Job — Prepare your honest answer, then use a six-part conversation outline, better phrasing and a physiotherapy clinic example to talk it through.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: AI receptionist prices as published by Smith.ai, Goodcall and Weave (September 2026) and summarised in the companion cost tutorial; staffing figures in the worked example are illustrative, not benchmarks.