Published answers to this question differ by a factor of fourteen.
Search “what does an AI employee save you” and you will get numbers between $50,000 and $700,000 a year for the same kind of business, most of them resting on a statistic that traces back to nothing. They cannot all be right. So here is ours in six steps — every input visible, every source named, the inputs other calculators skip left in, and the figure at the end split between the two AI Employees who would actually carry it.
Three things the total rests on.
None of them are ours to invent. Two are published findings you can go and read for yourself. The third is simply what the AI Employee does on the call, and we have not put a percentage on it.
- 01MIT Lead Response Management Study · 2007
Speed decides who gets the booking.
The one finding in this category with a retrievable source behind it. A study of more than 100,000 call attempts found that responding within five minutes made a business roughly 100 times more likely to make contact, and 21 times more likely to qualify the lead, than waiting thirty minutes. Thirty minutes — not thirty hours. Someone who reaches your voicemail at 5:15pm on a Friday is, by Monday, closer to a stranger than a lead. An AI Employee answers in seconds every time, which is the reason the pool above is treated as recoverable at all. What we will not do is convert that multiplier into a dollar figure — that is the second entry in the list of things this calculator refuses to price.
- 02Editorial estimate · 35%
Most of the week is unstaffed by arithmetic, not by fault.
A week is 168 hours. A front desk covers roughly 35 of them. That is not a criticism of your team, it is the shape of a working week, and it is why step 05 asks what share of your missed calls land outside opening hours. We default that input to 35% and label it an editorial estimate everywhere it appears, because no defensible study of after-hours call volume has been published — industry figures run from about a quarter to nearly half, and 35% is the midpoint of that range and nothing more. Change it to whatever your call log says. Note exactly what it does: it decides how much of the total sits with Tessa on evenings and weekends rather than with Ava during the day. It never changes the total itself.
- 03Product behaviour · not a statistic
A real answer comes before the ask.
A recovered call is only worth something if the conversation goes somewhere. Look at the sample exchange on the home page: the caller asks whether anyone is open tomorrow and gets two actual times back — 9:15am or 2:40pm — before being asked to commit to anything. That is the pattern across the roster. Kai answers the repetitive procedural and insurance-related enquiries that consume a receptionist’s morning; Ava gives real availability and writes the booking into the calendar. It is natural language, not a recording played at somebody, and it is not a form demanding a name and a number before it will tell you anything. We think that is why these calls behave like answered calls. We have not attached a conversion figure to it, because we do not have one worth quoting.
Measure your own miss rate first.
Every article on this topic, including ours, assumes you have a missed-call problem. You might not. Plenty of businesses have a front desk that genuinely answers the phone. Find out before you buy anything.
- 01Pull the call log from your phone system for the last 30 days. Most VoIP systems export this. You want total inbound calls, answered calls, and average ring duration.
- 02Divide unanswered by total. That is your real miss rate — not 38%, not 31%, yours.
- 03Check the timestamps. Missed calls cluster: lunch hour and Monday morning are the two worst windows in most businesses.
- 04Ask your front desk what happens to a voicemail on a Saturday.
If your miss rate is under 10%, none of this applies to you. Close the tab — you have spent ten minutes and confirmed your front desk is doing its job.
Every default above, and where it came from.
Six entries. Five trace to a published dataset you can go and read. The sixth is an editorial estimate and is labelled as one, here and in the calculator itself. Where we could not find a source we were willing to stand behind, there is no number at all rather than a borrowed one — there are three of those, and they are listed underneath. Several of the defaults come from dental-industry research, because that is where the best public data on inbound-call economics happens to live. None of them are a claim about your business, whichever of the eight industries you are in — every one is an editable input, and the point of the six steps is that you replace ours with yours.
31–38% of inbound calls go unanswered
Peerlogic (4,280 calls / 26 practices, Feb 2026) and Patient Prism (11.5M calls / 8,280 locations, 2025)
Both vendors sell call software and sample practices that already installed call tracking. Best available, not independent audits.
Roughly 30% of inbound calls are new-client calls
The input most published figures skip — which is what inflates the headline numbers
A missed call is not the same thing as a missed new client.
New-client calls book 21–25% of the time
Patient Prism (21 bookings per 100 calls) and Peerlogic (25.2%) — both dental call-tracking datasets
Even answered calls book about a quarter of the time. Whatever your industry, replace this with your own rate.
$685 average annual per-customer value; $4,500–$6,500 lifetime
American Dental Association Health Policy Institute, Medical Expenditure Panel Survey — dental-industry data, the best published figure of its kind we could find
It is a default, not a claim about your business: every industry on this site has its own number, and this input exists so you can put yours in. The widely-quoted "$300–$600 lifetime value" figure is first-visit production, not lifetime value.
Responding within five minutes makes contact ~100x more likely
MIT Lead Response Management Study, Dr. James Oldroyd (MIT Sloan School of Management, with InsideSales.com), 2007 (100,000+ call attempts)
Also 21x more likely to qualify the lead, versus waiting 30 minutes. Deliberately not converted into a dollar figure anywhere on this page — see below.
Roughly a third of missed calls arrive outside business hoursEditorial estimate
No defensible published study exists. Industry estimates put after-hours volume between a quarter and nearly half of inbound calls; 35% is the midpoint of that range and nothing more.
This is the one input on this page we cannot source. It is an editorial estimate, it is yours to change, and it only decides how the total splits between two AI Employees — it never changes the total itself.
Three things this calculator refuses to price
Every one of these would make the total on this page larger, and we have no honest way to calculate any of them. A blank is worth more to you than a number we made up.
A reactivation figure for Grant
Grant works dormant contacts already sitting in your database. Putting a number on that needs two things we do not have: how many dormant records you hold, and what share of them rebook when worked. The first is yours, not ours. The second has no published figure we can defend. So there is no third bucket here, and Grant is quoted at zero rather than at a guess.
A dollar value on speed-to-lead
The MIT Lead Response Management Study finding — 100x more likely to make contact inside five minutes — is real and cited above. Turning a contact-rate multiplier into revenue needs your current baseline contact rate, which is not one of the inputs on this page. Competing calculators bridge that gap with an invented "up to 20% conversion lift". We leave the multiplier as what it is: a reason, not a line item.
A time-saved metric
Minutes saved per handled call is the easiest number in this category to invent and the hardest to source. We found no defensible per-interaction figure, so there is no hours-reclaimed readout here. If you want one, time ten of your own calls.
What people ask after running it.
The general ones — pricing, compliance, what happens to your agency — are answered on the home page.
No, and nothing on this page should be read as one. It is arithmetic performed on six inputs, five of which you can and should replace with your own. What is guaranteed is published separately and stated plainly: we guarantee your AI Employees will generate 3x more appointments within 90 days, or we will personally optimise it until they do. That is a commitment about appointments booked. It is not a promise that the figure above will land in your account.
Then one of the inputs was wrong, and the honest thing is to find out which. That is why every step shows its own arithmetic rather than one opaque total — your call volume, new-client mix, booking rate and client value are each separately checkable against your own systems at ninety days. On the delivery side, the systems are monitored and there are human oversight tools in place, so a problem gets caught and corrected quickly rather than repeating quietly for weeks.
Compare the two numbers directly, because both are published. One AI Employee is $3,000 one-time setup, four together are $10,000 under the Go All In plan, and support and updates run on a subscription. Setup fees are non-refundable. Put your own miss rate into step 01 and see what the first-year figure comes to. If it does not clear the setup fee with room to spare, do not buy it — and if your miss rate is under 10%, we would rather you closed the tab.
Three differences, and none of them are about the technology underneath. It is trained on your business, your scripts and your objections, and tested against your real call and enquiry scenarios before it goes live — it is not a generic bot answering from a script somebody else wrote. It takes actions rather than only holding a conversation: it connects to your CRM, calendar and phone system and writes the booking into the schedule itself. And it is hired for one job across the channels the enquiry actually arrives on — voice, SMS and email, 24 hours a day — rather than waiting on a page for someone who already found you. Ava books appointments. Grant works your dormant database. Each one is scoped to a real job description, not a feature list.
Yes. The goal is enhancing your team, not replacing it. An AI Employee handles the overflow, the after-hours calls and the repetitive scheduling work that eats your team’s day, so the people you employ spend theirs on the things that genuinely need a person — including the client standing in front of them.
It hands over rather than improvises. Complex insurance questions and anything clinical still need a human, and we would rather tell you that now than after you have signed something. Kai handles the repetitive procedural and insurance-related enquiries that consume a receptionist’s morning, and anything past that goes to your team instead of being guessed at.
A 20-minute kickoff call, plus roughly 30 minutes of homework from you or a member of staff. That is the whole commitment on your side. Integration with your CRM, calendar and phone system, and testing against your real call and enquiry scenarios, are part of the setup fee rather than a separate project.
Because we would have to invent it. Calculators of this kind routinely apply a conversion lift to their total, and the figure has no retrievable source. The speed finding underneath it is real and cited on this page, but turning a contact-rate multiplier into revenue needs your current baseline contact rate, which is not one of the inputs here. So the multiplier stays what it is — a reason the recovered calls convert at all, not a line item that inflates the total.
Because it answers a different question. First-year value is what one new client spends in twelve months; $685 is the American Dental Association Health Policy Institute average and your own software has yours. Lifetime value spans seven to ten years of retention, which puts it between $4,500 and $6,500. One year of missed calls therefore produces a five-figure first-year number and a six-figure lifetime one, and both are true — the six-figure version is the one most articles quote without saying which they mean. Only the first-year figure feeds the split between Ava and Tessa.
The same way you should measure the problem before you start. Export the last 30 days from your phone system and record your miss rate, your new-client mix and your booking rate now, then do it again at ninety days. Those three numbers, from your own system, settle the question better than anything on this page. If you want a time-saved figure as well, time ten of your own calls — we do not publish a per-interaction number because we could not find one worth defending.