Case Study · Practice ROI
An AI dental receptionist on every phone, in five offices.
Here is month one.
A dental group with five offices ran an AI receptionist on every line for a full month. This is not a testimonial and a round number. It is the actual call logs, with the math shown, and then checked the next month.
By Fortify Practice · Published September 2026 · 9 min read
$57,788
Net return after cost
675%
Combined ROI
573
Calls handled
$66,342
Production booked
Here is the short version.
A dental group with five offices put an AI receptionist on every phone for one month. It answered the calls the front desk could not get to, booked those patients into the schedule, and Fortify tracked every dollar it produced.
The result: 573 calls handled, $66,342 in production, and a 675% return on what the AI cost. Every one of the five offices made money. Below is the whole thing, explained in plain English, with the charts.
New to AI receptionists? Read this first.
An AI receptionist is software that answers your phone in a real, natural human voice, has a normal conversation with the patient, and books the appointment straight into your scheduling software. It works the calls your team cannot reach: the lunch rush, after hours, weekends, and every time all your lines ring at once. In this case study, that AI is Lucy.
What this means for one practice
This report is five offices. If you run one practice, here is the number to hold onto:
$13K
recovered production per month, on average
13x
return on what the AI costs
2
new patient visits to fully cover the monthly cost
Your number depends on your call volume and how many calls you miss today. Run your own in about 30 seconds.
Key takeaways
- Across five offices in one month, an AI receptionist handled 573 calls and booked $66,342 in production for an $8,554 cost. That is a 675% return.
- Every one of the five offices finished positive, from 113% up to more than 1,600% ROI.
- Production is counted only when a patient actually showed up and paid, pulled from the practice software, not when an appointment was merely booked.
- The AI did not replace the front desk. It caught the overflow and after hours calls the team could not reach.
- It was not a fluke. The next month, the same group grew to $120,068 in production.
1
The setup: what we did and what we measured
This was the first full month with an AI receptionist live across all five of the group's offices at the same time.
The rule was simple. Any inbound call the front desk did not pick up, during the lunch rush, on the other line, after hours, on a weekend, rang through to the AI. It answered, had a real conversation, and booked the appointment straight into the schedule before the caller hung up.
Why does that matter? Because of what dental phones actually do all day.
- Dental practices miss somewhere between 20% and 38% of incoming calls during business hours.
- Roughly 80% of callers who reach voicemail never leave a message. They just call the next practice.
The group was not losing those calls because the front desk was bad at its job. It was losing them because one person cannot answer three ringing phones at once.
So read this report the right way. It is not AI versus your front desk. It is your front desk plus the calls your front desk could never physically get to.
Three words we use a lot, in plain English
- Production: the money a practice actually collects for the work it does.
- Show rate: the share of booked patients who actually showed up.
- Scheduling call: a call from someone really trying to book, reschedule, or get in for an emergency (not a confirmation, a bill question, or a vendor).
2
The results, by the numbers
Of the 573 calls the AI touched, here is how the work split between the AI and the team, and the production each slice booked.
- AI fully handled214 calls$37,226 booked
- AI, then a rep76 calls$15,853 booked
- Rep took over180 calls$7,801 booked
- Not logged103 callsunclassified in the log
Notice the honest bit. The AI handled 214 calls entirely on its own, and it booked the most production of any slice. But it did not do everything. On 180 calls, a human took over, exactly where a person was the better fit. That is the point.
One more number keeps the rest honest. Of the 573 total calls, only 307 were true scheduling calls. The other 266 were confirmations, billing questions, and vendor calls that were never going to become an appointment. So booking performance should be judged against those 307, not the full 573.
Key takeaway The AI booked real money on the calls the office was already missing, and it handed the rest to a human. Front desk plus AI, not front desk versus AI.
3
Which office won, and by how much
Averages hide things. So here is every office, ranked by the production the AI booked, with each office's return and its call volume. Every single one finished the month positive.
Read that chart with your own office in mind. The biggest office produced $29,765 from 204 calls. But look at the smallest one: just 62 calls, and it still returned 113%, more than doubling its money, while the AI handled six of every ten of its calls alone. Two offices posted a perfect 100% show rate for the month.
Key takeaway No single office carried the group. Big or small, busy or quiet, all five made money in month one.
4
Where the calls came from
The AI was not just answering the main line. The month's booked patients came from four places, and the phone was the engine.
Here is a nice bonus. On top of the phones, automatic follow up texts went to patients who never called back on their own. Those texts had a booking link, and 41 patients booked themselves, no phone call at all. That is a second lane of recovery running quietly in the background.
5
Why it worked (three plain reasons)
Strip away the dashboards and the return came from three unglamorous things.
It answered the calls that were already being lost
Nothing here needed new marketing or more patients calling. The calls were already coming in. The AI just stopped letting the overflow and after hours ones ring out. That is the cheapest production in a practice, because you already paid to make the phone ring.
It booked on the call, not later
A message or a callback adds the delay that loses patients. Here, the appointment landed on the schedule while the caller was still on the line. That is why the booking rates on scheduling calls ran from the high 70s into the 90s.
It worked with the team, not around it
The front desks did not shrink. The AI took the calls they could not reach and handed off the ones that needed a person. You are not replacing salaries. You are recovering the production leaking between the calls your team can physically answer. That is the same AI versus a front desk hire math, proven with real logs.
6
How we measured this (the honest part)
Fair question: how do you know the AI caused this money, instead of just being in the room when it happened? Here is exactly how the numbers were counted, so you can trust them.
What counts as an AI booking?
Only appointments the AI actually booked on the call. If a human booked it, it does not count for the AI.
When is production counted?
When the patient shows up and pays, pulled from the practice software. Not when the appointment is booked. A no show produces nothing, so no shows are not counted as production.
Why measure against 307 calls, not 573?
Because 266 of the calls (confirmations, billing, vendors) were never appointment chances. Judging booking rate against them would flatter the number. We judged it against the 307 real scheduling calls instead.
What is the "103 not logged"?
Calls that were not cleanly classified in the log for the month. They are left out of production entirely rather than guessed at. That makes the results more conservative, not less.
What did the $8,554 cost include?
The total AI cost for the month across all five offices, subscription plus per minute usage. Standard Lucy pricing is $997 a month per practice, a one time $1,600 setup, and usage.
Bottom line These are booked and showed dollars, credited only to calls the AI actually handled, measured against the calls that could really have become appointments.
7
"But will patients hate it?" and other fair questions
If you are skeptical of AI, good. Here are the objections we hear most, answered straight.
Won't my patients hate talking to a robot?
Patients care about getting help, not about who picks up. Lucy answers right away, holds a natural conversation, handles the reason for the call, and books the appointment instead of sending them to voicemail. When a call needs a person, the team steps in.
Is this even HIPAA compliant?
Yes. Patient information is handled under a signed Business Associate Agreement. The AI only uses it to answer the call and book the appointment for your practice, under your instructions.
What about the calls it got wrong?
When the AI hits a question it is not trained for, it does not guess. It takes a message, tells the patient someone will follow up, and flags it for the team. On 180 calls this month, a human stepped in. That is the design, not a failure.
Isn't a 675% return too good to be true?
It is the combined number for a busy five office group, and it is on the high side. Individual offices in the very same report ranged from 113% to over 1,600%. Your number will land somewhere in your own range, which is why we show the whole spread instead of one hero figure.
How much of my time and my staff's time does this take?
Setup runs about two weeks, and Fortify does the setup, not you. Your team keeps working the way they already do. The AI simply answers the calls they cannot reach.
8
Was it a one month fluke? No.
One good month can be luck. So here is the real test: what happened the next month, with the same group and the same offices.
Production grew 81% month over month, and the return nearly doubled. It was not a spike. As the AI settled in and the follow up system got going, it recovered more, not less.
9
What this means for your practice
Time for honesty before any pitch. This is one group's results. Your ROI depends on your call volume, how many calls you miss today, and your average production per patient. Treat 675% as an illustration, not a guarantee. Offices in this very report returned as little as 113%.
What is consistent is the shape of it. Across the practices Fortify works with, the average is about $13K a month in recovered production, roughly a 13x return on what the AI costs. It holds because every practice has the same leak: calls that ring out during the rush and after hours, from patients who then call someone else. Close that leak and the production comes back.
Two new patient visits a month cover the entire cost. Everything above that is production you were already losing.
See the month for your own practice.
Not ready for a sales call? Start smaller. Hear Lucy answer a real patient call, or run your own missed-call math first.
Client details and office names anonymized. Figures are the group's actual reported performance across all five locations.