# How to Calculate ROI on an AI Receptionist

> Every vendor ships a calculator pre-filled with its own assumptions. Here is the arithmetic that uses your numbers instead, including the discount none of those calculators apply.

**Author:** gethelpdesk-ai  
**Published:** 2026-09-15  
**Category:** guides  
**Tags:** ROI, pricing, evaluation, procurement, missed-calls, dental-practice

**Canonical:** https://gethelpdesk.ai/blog/calculate-roi-ai-receptionist-dental/

---

## Every Calculator Agrees With the Vendor That Built It

Open any AI receptionist website and you will find a slider. Move it, and a number appears telling you what your missed calls are costing you. The number is always large, and it is always larger than the price of the product on the same page.

Two examples, both read on 15 September 2026, both from the vendor's own homepage. Dentina's calculator opens pre-filled and returns roughly $55,440 a month at risk, with its inputs labelled "industry avg. ~33%" for missed new-patient inquiries and "first-year production, avg. $4,200" for the value of a new patient. Viva AI's opens at about $23,382 a month at risk, and adds that it "typically recovers roughly $14,029 a month of that, booking about 60 percent of answered calls," which it then describes as 16 times the cost of its $899 Platinum plan.

Neither number is dishonest. Both are clearly presented as estimates, and Viva's is explicitly labelled as back-of-napkin math rather than a quote. But neither is a measurement of your practice, and you cannot take either into a partners' meeting.

The method below takes about twenty minutes, uses five numbers you already have, and produces a figure you can defend when someone asks where it came from.

One of them is the cost side, and the monthly figure plus what belongs inside it is in [what a dental AI receptionist costs](/blog/what-does-a-dental-ai-receptionist-cost-monthly/).

## The Five Numbers That Are Actually Yours

Four of these are in your phone system's reporting. The fifth is in your practice management system.

1. **Inbound calls per month.** Total, not new patients. Your phone provider reports this.
2. **The share that go unanswered.** Rung out, hung up in queue, or landed in voicemail. Count abandoned calls, not just voicemails left.
3. **The share of missed calls that are new patients.** If your system does not tag this, sample two weeks of unknown numbers by hand. It is worth the hour.
4. **Your booking rate on answered new-patient calls.** Your front desk already knows this one, and it is usually higher than people guess.
5. **Your own first-year production per new patient.** Not an industry average. Pull your last fifty new patients and take the median, not the mean, so a single full-arch case does not distort it.

If you only ever do one thing from this article, do number five. Every vendor calculator ships with a national-sounding figure in that field, and it is the input that swings the result hardest.

## Step One: The Revenue Side

Here is the arithmetic, worked on a deliberately ordinary practice. Substitute your own numbers as you go.

| Input | Example | Yours |
|---|---|---|
| Inbound calls per month | 400 | |
| Share unanswered | 18 percent | |
| Missed calls per month | 72 | |
| Share of missed calls that are new patients | 25 percent | |
| Missed new-patient calls | 18 | |
| Booking rate when answered | 50 percent | |
| Newly booked patients per month | 9 | |
| First-year production per new patient | $1,200 | |
| Gross monthly opportunity | $10,800 | |

We used $1,200 rather than the $4,200 that Dentina's calculator assumes, because a conservative input is the one that survives being questioned. If your own median is genuinely higher, use it, and be ready to show the fifty charts it came from.

## The Discount No Vendor Calculator Applies

Here is the part that matters, and the reason every published calculator overstates its result.

Not every missed new-patient call is a lost patient. Some of those callers ring back. Some leave a voicemail your team returns at 8:30 the next morning and books. Some had already found you through a referral and were always going to reach you. A calculator that counts every missed call as lost revenue is counting patients you would have got anyway.

Nobody can tell you the right discount for your practice, but you can estimate it honestly. Pull thirty missed calls from last month and check how many of those numbers appear in your schedule within the following fortnight. That percentage is your callback recovery rate, and it is the share you should strip out.

On the example above, assume half of those missed new-patient callers would have reached you eventually:

- 9 newly booked patients per month, minus 50 percent already recovered
- **4.5 genuinely incremental new patients per month**
- 4.5 multiplied by $1,200 equals **$5,400 in incremental monthly production**

That is less than half the gross figure, and it is the number to use. Note that the conclusion does not change: $5,400 a month still dwarfs any price in this category. The point of the discount is not pessimism. It is that a defensible number and a flattering number reach the same decision here, and only one of them survives scrutiny.

## Step Two: The Cost Side, Normalised

This is where comparison actually breaks down, because the category prices itself three incompatible ways: flat monthly, per credit or minute, and per call. A headline price tells you nothing until you convert it.

All figures below are from each vendor's own published pages, read on 15 September 2026.

| Vendor | Published model | What it costs at the entry tier |
|---|---|---|
| GetHelpdesk.AI | Flat monthly, unlimited calls | $399 a month, all providers, plus a one-time $500 setup fee, first month free |
| Dentina | Flat monthly, unlimited calls and minutes | Standard Inbound from $299 a month, billed annually, per location, up to 3 providers, no setup fee |
| Viva AI | Monthly credit allotment, then overage | Gold $349 a month for 4,000 credits, additional credits $0.10 each |
| Smith.ai | Per call | Pro $150 a month at $2.00 a call, 75 calls |
| Arini | Publishes no pricing | Not published |

### The derivations nobody shows you

**Viva AI.** Its own page states 10 credits per minute of AI-answered call, 5 credits per AI-answered message, and 1 credit per Office Optimizer question. So Gold's 4,000 credits is **400 minutes of answered calls a month**, and $349 divided by 400 is about **$0.87 a minute**. Past that, $0.10 a credit is **$1.00 a minute**. Platinum's 7,000 credits is 700 minutes at about $1.28 a minute, with overage at $0.90. Diamond's 12,000 is 1,200 minutes at about $1.00 a minute, with overage at $0.80. Viva states that 95 percent of offices never exceed their allotment. Note that messaging draws on the same pool, so 400 minutes is a ceiling rather than a floor.

**Smith.ai.** Per-call pricing cannot be converted to a per-minute rate without your own average call length, and that cuts in your favour if your calls run long. At $2.00 a call, a three-minute average is about $0.67 a minute and a six-minute average is about $0.33. Its published tiers are internally consistent: 75 calls at $2.00 is the $150 Pro plan, and 300 calls at $1.67 is the $500 Enterprise plan. Smith.ai is a horizontal answering service rather than a dental specialist, which is a separate question from price.

**Flat and unlimited.** Dentina and we both publish unlimited calls, so the per-minute cost simply falls as your volume rises and the floor is the monthly fee. This is the model that does not punish a busy month, and it is the only one where a marketing campaign does not also raise your phone bill.

## Step Three: Year One Is Not Steady State

Setup fees, free months and annual commitments all land in the first twelve months and then disappear. Compare both.

| | Year one | Every year after |
|---|---|---|
| GetHelpdesk.AI | $500 setup plus 11 months at $399 equals **$4,889** | **$4,788** |
| Dentina Standard | 12 months at $299, billed annually, no setup fee, equals **$3,588** | $3,588 |
| Dentina Premium | 12 months at $399, billed annually, no setup fee, equals **$4,788** | $4,788 |
| Viva AI Gold | 12 months at $349 equals **$4,188**, if you stay inside 400 minutes | $4,188 |
| Smith.ai Pro | 12 months at $150 at 75 calls equals **$1,800** | $1,800 |

Two honest observations about our own column. Dentina's Standard tier is $1,301 cheaper than us in year one, and its 30-day free trial and absence of a setup fee mean a small practice can hear it answer a real call before spending anything. Our $500 is payable up front. Against Dentina Premium the two are within about $100 in year one and identical afterwards, but theirs is billed annually and ours is month to month, which is a real difference in what you are committing to.

The figures that look cheapest also carry limits worth reading. Dentina's tiers are per location and cover up to three providers. Viva's Gold allotment is 400 minutes. Smith.ai's $150 covers 75 calls, which is under three a day.

## The Payback Line

Put the two halves together. On the example practice, at $5,400 a month in genuinely incremental production against $399 a month:

- **Payback on the monthly fee: about two days of the month.**
- **Payback including the $500 setup: still inside the first month**, and the first month is free, so the setup fee is really recovered in month two.
- **Break-even volume: one incremental new patient roughly every three months**, at $1,200 each, covers the annual cost.

That last line is the one to carry into the partners' meeting, because it inverts the question. You are no longer asking whether the projection is right. You are asking whether the service will recover four new patients a year that you would otherwise have lost. If the answer to that is obviously yes, the precision of the rest stops mattering.

## What Would Actually Make This a Bad Investment

An honest model should be able to produce a negative answer. Three ways it does:

- **You do not actually miss many calls.** If your unanswered rate is 3 percent and your team returns voicemails within the hour, the recoverable pool is small and the case is weak. Measure before you model, and [measure again afterwards](/blog/audit-ai-receptionist-booking-conversion/) to check the model held.
- **It answers but does not book.** A system that captures a message and emails it to the front desk has moved the work rather than removed it. The revenue side of this model assumes appointments land in your practice management system. If it cannot write into yours, delete the revenue side.
- **It books badly.** Double-booked operatories, appointments outside your block templates, and duplicate patient records all carry a real cost that never appears in a vendor calculator. We wrote about how that fails in practice in our guide to [setting up AI call answering in Eaglesoft](https://gethelpdesk.ai/blog/gethelpdesk-ai-eaglesoft-integration-guide/).

## FAQs

**What is a realistic payback period for an AI receptionist?**
For most practices with a measurable missed-call rate, under one month on the monthly fee. That sounds like marketing, but it falls out of the arithmetic: monthly cost in this category runs from $150 to $1,200, and a single recovered new patient in general dentistry usually exceeds that. The real question is not the payback period, it is whether the missed calls exist in the first place.

**Should I use the vendor's calculator at all?**
Use it to understand the vendor's assumptions, which is genuinely informative, then rebuild it with your own five numbers. The input that matters most is first-year production per new patient, and no vendor knows yours.

**How do I compare a per-call price against a flat fee?**
Multiply the per-call rate by your actual monthly call volume, not by the volume in their example tier. At 400 calls a month, a $2.00 per-call rate is $800, which is more than double a $399 flat fee. At 75 calls a month it is $150, which is far less. Your volume decides it, and it is a number you already have.

**Should I count staff time saved as part of the return?**
Only if you will actually do something with the time. Hours freed at the front desk are real, but they do not become money unless they turn into treatment presentation, recall follow-up, or a role you do not have to backfill. Count it separately from the revenue line and label it honestly.

**Does the return change once we are past year one?**
Yes, and usually in your favour, because setup fees and onboarding effort are one-time while the recovered calls are not. Run both columns before you sign, and be wary of a first-year figure that only works because of a discount that expires.

## Where to Start

Pull your five numbers before you book a single demo. The exercise takes an afternoon and it changes every sales conversation afterwards, because you stop being shown a slider and start asking vendors to price against your actual volume.

Then, when you have a shortlist, run the practical tests in our [buyer's guide](https://gethelpdesk.ai/blog/ai-receptionist-buyers-guide-dentists/). If you want the shape of the decision rather than the money, start with [which AI receptionist is best for your practice](https://gethelpdesk.ai/blog/which-ai-receptionist-is-best-for-your-practice/).

## Sources

Vendor pricing figures were read on 15 September 2026 from each vendor's own published pages: Viva AI's homepage pricing section and credit schedule, Dentina's homepage pricing section and calculator, and Smith.ai's AI receptionist pricing page. Arini's homepage was checked on the same date and publishes no pricing figure and no pricing page. Vendor marketing pages change without notice, and the figures above are accurate only as of that date. GetHelpdesk.AI figures come from our own [pricing page](https://gethelpdesk.ai/pricing/). The worked example is illustrative and uses no practice's real data.

---

**Related reading:** the inputs to this calculation should come from your own systems. Start by working out [find your practice's real missed-call number](/blog/find-your-real-missed-call-number/).