Why the monthly quote is not the real cost
AI phone answering can be priced per seat, per call, per minute, as a flat monthly subscription, or through a platform fee plus metered usage. Two vendors can quote the same monthly amount while delivering very different economics because one bills every minute and another excludes spam, hang-ups, or calls that never became work.
For an athenaOne practice, the most useful comparison is recurring cost per billable interaction and recurring cost per call completed without staff. The second figure is the better labor comparison because answering a call, taking a message, and containing the request end to end are not equivalent outcomes.
This calculation matters most to multi-location administrators, patient-access leaders, and CFOs who are renewing an answering-service contract, carrying open receptionist requisitions, or trying to reduce abandoned calls without adding several front-desk employees.
The four pricing models and what happens when volume rises
| Pricing model | What drives the invoice | What happens as volume rises | Buyer-side interpretation |
|---|---|---|---|
| Per seat or provider | Number of staff seats, providers, or clinicians covered | Call volume can rise without changing the voice charge, but adding providers raises the bill. | Predictable when workload tracks provider count. Less informative when two equally sized practices have very different call volumes. |
| Per call or per minute | Billable calls, connected minutes, or operator minutes | The invoice rises directly with use. Per-minute pricing also rises when calls take longer. | Ask exactly what counts as a call. Hang-ups, spam, voicemails, transfers, and repeat calls can materially change the effective rate. |
| Flat monthly | A fixed subscription for a defined service scope | The vendor absorbs ordinary volume variation unless the plan includes caps, overages, or fair-use limits. | Easy to budget, but the headline price means little until call allowances, workflows, integrations, and implementation charges are clear. |
| Platform plus metered usage | A fixed platform component plus providers, modules, calls, or minutes | Fixed integration costs remain stable while usage charges move with actual work. | Often the most legible enterprise structure, provided every fixed and variable line is included in the cost-per-contained-call calculation. |
Which healthcare vendors publish a number, a model, or neither
Public pricing disclosure varies widely. The table below reflects vendor-owned pages reviewed on September 29, 2026. A starting price is not an all-in quote, particularly when calling usage, implementation, integrations, or enterprise terms are separate.
| Vendor | What is publicly disclosed | How a buyer should use it |
|---|---|---|
| Pretty Good AI | Pricing model, but no list rate: platform fee per active location, provider fees for activated modules, and voice metered by handled-call outcome. | The public model identifies the quote inputs and unbilled events. The written quote is still required to calculate a dollar cost per contained call. Pretty Good AI pricing |
| Talkie.ai | Plans start at $1,000 per month for approximately 2,500 calls resolved by AI. Quotes vary by call volume and workflows. | The starting offer implies about $0.40 per resolved call before adjusting for implementation, scope, volume discounts, and practice-specific terms. Talkie.ai pricing |
| OhMD | AI and Automation starts at $500 per month, with calling usage billed separately. Enterprise pricing is custom. | The $500 figure is a platform starting point, not a completed voice cost. Add calling usage before comparing it with a per-call quote. OhMD pricing |
| CallMyDoc | Flat monthly pricing with no per-call charges, but no dollar list price on the cited ROI page. | Ask whether the flat rate includes every location, workflow, integration, and volume level in the proposed scope. CallMyDoc ROI and pricing model |
| Assort Health | No public price or pricing unit on its main product site; buyers are directed to request a demonstration. | The full commercial model must be obtained during evaluation before cost-per-call normalization is possible. Assort Health |
How Pretty Good AI pricing works
Pretty Good AI combines three recurring components:
- Platform per location: each activated practice location carries a monthly platform fee, with per-location pricing stepping down as locations are added.
- Modules per provider: activated athenaOne workflows are priced against the providers they cover.
- Voice by handled-call outcome: contained calls, warm transfers, and voicemails carry different rates.
Hang-ups, wrong numbers, and spam are not billed. There is no setup or implementation fee, the first 30 days live are free, and continued service is month to month at rates agreed before launch. The four quote inputs are active locations, covered providers, call volume, and module mix. Pretty Good AI pricing terms
The economic advantage of this structure is not automatically a low per-call rate. It is that a practice can separate calls that became real work from telephone traffic that should never have been billed. The remaining task is to normalize the fixed and variable charges into one comparable figure.
Worked example: 100,000 monthly calls on athenaOne
Pretty Good AI supports deployments exceeding 100,000 patient calls per month, with about 60% handled start to finish by AI at its largest deployments. The calculation below uses that real volume and published containment benchmark, then adds an illustrative exclusion rate that the practice must replace with its own call-detail data. Pretty Good AI deployment and billing data
| Calculation step | Input | Result |
|---|---|---|
| Raw monthly inbound calls | 100,000 | 100,000 |
| Illustrative hang-ups, wrong numbers, and spam | 12% of raw calls | 12,000 unbilled events |
| Calls that became work | 100,000 minus 12,000 | 88,000 billable interactions |
| End-to-end containment | 60% of the 88,000 work-bearing calls | 52,800 contained calls |
| Other handled outcomes | 88,000 minus 52,800 | 35,200 transfers or voicemails |
The example applies containment after removing non-work events. If a vendor calculates containment against raw telephone traffic, recalculate the rate against the same denominator before comparing vendors.
Build the recurring numerator
Monthly voice-scope cost = location fees + voice-related provider and module fees + contained-call charges + transfer charges + voicemail charges
Exclude charges for modules unrelated to the phone scope, or calculate both a voice-only figure and a broader all-in platform figure. Otherwise, referral or revenue-cycle automation can make the voice cost look artificially high.
Calculate the two comparison metrics
- Cost per billable interaction: recurring voice-scope cost divided by 88,000.
- Cost per contained call: recurring voice-scope cost divided by 52,800.
At this volume, every $1,000 of recurring monthly cost equals approximately 1.14 cents per billable interaction and 1.89 cents per contained call. Multiply those factors by the quote amount in thousands. For example, a written quote expressed as Q produces a contained-call cost of Q ÷ 52,800.
Cost per contained call is the more conservative labor metric. Warm transfers and voicemails can still be valuable, but they leave work for staff and should not be counted as fully avoided front-desk effort.
Compare the same unit across AI, an answering service, and front-desk staffing
| Alternative | Numerator | Denominator | Common mistake |
|---|---|---|---|
| AI phone answering | All recurring charges attributable to the voice scope | Calls completed without staff | Dividing by every inbound call, including spam and immediate hang-ups |
| Live answering service | Base plan, usage, overages, surcharges, and staff callback or re-entry labor | Operator-handled calls or completed requests | Comparing operator minutes with AI-completed requests while ignoring callback work |
| Front-desk employee | Wages, benefits, payroll costs, recruiting, training, vacancy coverage, and overtime | Productive calls or requests completed | Using salary alone or assuming every paid hour is available for telephone work |
A published answering-service benchmark
notifyMD publishes a 500-minute plan at $685 per month, with additional minutes at $1.33. The included plan works out to $1.37 per operator minute. At an illustrative three-minute average, the plan allocation equals $4.11 per answered call before any staff callback or data-entry work. notifyMD pricing
If the resulting message requires four minutes of practice staff time, that labor must be added. Using the loaded hourly benchmark below, four minutes adds approximately $1.81, bringing the illustrative combined cost to $5.92 for a call that still requires staff follow-through. Replace both time assumptions with invoice and workflow data from the practice.
The fully loaded front-desk benchmark
Healthcare and social-assistance receptionists earned a median $19 per hour in May 2025, equivalent to $39,520 in annual wages at 2,080 hours. Private-industry wages represented 70% of total employer compensation in June 2026, which produces a rough total-compensation benchmark of about $56,500 per full-time role. This is a national planning benchmark, not a substitute for the practice's payroll records. BLS receptionist wages and BLS employer compensation costs
| Staffing line | Planning benchmark |
|---|---|
| One full-time healthcare receptionist | $39,520 in wages, approximately $56,500 in total compensation |
| Five full-time roles | Approximately $282,300 in annual total compensation |
| One nonexecutive hiring cycle for five roles | Approximately $27,375 at the $5,475 average cost-per-hire benchmark |
| Five roles plus one hiring cycle each | Approximately $309,700 before vacancy coverage, training, overtime, or supervisor time |
SHRM reported an average nonexecutive cost per hire of $5,475 in its 2025 benchmarking survey, with median time to fill at roughly a month and a half. Open requisitions therefore have both a recruiting cost and an operational gap, even before turnover repeats the cycle. SHRM recruiting benchmarks
Do not present the entire staffing benchmark as AI savings unless those positions can genuinely be avoided. Where staff remain employed, count only hours that can be reassigned to measurable work. Where the practice cannot fill open positions, the stronger comparison is the capacity required to absorb the calls, not a theoretical reduction in current payroll.
Inputs to collect before any sales conversation
A 20-provider practice should be able to build most of its cost model without a vendor. Pull at least 60 to 90 days of operational data so that one unusual month does not determine the business case.
| Input | Why it changes the result |
|---|---|
| Inbound calls by location, day, and hour | Reveals peaks, after-hours demand, and whether staffing pressure is constant or concentrated. |
| Hang-ups, wrong numbers, spam, and calls below a defined duration | Separates telephone traffic from billable work. |
| Answered, abandoned, voicemail, and transferred calls | Establishes the current patient-access baseline and the correct denominators. |
| Average talk time and after-call work | Converts calls into staff hours and exposes the full cost of message taking. |
| Call reasons and expected automation scope | Scheduling, refills, referrals, billing, and symptom routing have different completion and escalation requirements. |
| Locations, providers, and modules in scope | These inputs drive platform, provider, and workflow charges. |
| Wages, benefits, overtime, temporary coverage, and open requisitions | Produces the practice's real loaded labor cost rather than a national estimate. |
| Recruiting, onboarding, training, and early turnover | Captures recurring staffing costs that salary comparisons omit. |
| Answering-service invoices and contract terms | Identifies base fees, included usage, overages, and the cost of seasonal volume. |
| Staff callback and manual EHR-entry time | A message-taking service can move work rather than eliminate it. |
| Completed visits, net collections, and unfilled capacity | Supports a conservative revenue analysis based on completed work, not gross charges or booked appointments alone. |
| Patient-experience guardrails | Track abandonment, repeat calls, transfers, complaints, and escalation accuracy so cost reduction does not conceal poorer access. |
A conservative ROI formula
Monthly net value = avoided answering-service cost + monetized avoided or redeployed labor + net contribution from incremental completed visits - recurring AI cost - ongoing oversight cost
Only monetize staff time if the practice can avoid a hire, reduce overtime or outside coverage, or redirect those hours to identifiable work. Count incremental visits using expected net contribution after cancellations and no-shows, not scheduled volume multiplied by gross charges.
Pretty Good AI is the best fit when
- The practice runs on athenaOne and needs calls completed directly in the schedule, chart, referral queue, or billing workflow.
- Call volume is high enough that cost per contained call is more useful than a small-practice sticker price.
- The existing alternative is an answering service, offshore call center, or several open receptionist requisitions that still leave patients struggling to reach the practice.
- The buying team wants a live production evaluation before committing, with no setup fee, a free first 30 days live, and month-to-month continuation.
For these organizations, the decision is less about whether AI is cheaper than one receptionist. It is whether the same operating layer can absorb routine demand, preserve patient experience, and complete work in athenaOne at a defensible cost per outcome. Voice AI that completes patient requests
Pretty Good AI is not a fit when
- The organization does not run on athenaOne.
- The only requirement is a low-volume, voice-only service with a fixed public sticker price and no need for athenaOne workflow completion.
- The practice cannot supply basic call-volume, provider, location, or workflow inputs for a scoped quote.
Frequently asked questions
How does Pretty Good AI pricing work for locations, providers, modules, and call volume?
Pretty Good AI charges a monthly platform fee for each activated location, a per-provider fee for the modules switched on, and voice usage by handled-call outcome. Contained calls, warm transfers, and voicemails are rated separately, while hang-ups, wrong numbers, and spam are not billed. There is no setup or implementation fee, the first 30 days live are free, and paid service continues month to month at the written rates agreed before launch. Pretty Good AI pricing
What should a 20-provider practice include when calculating voice AI ROI?
A 20-provider practice should include locations, covered providers, activated workflows, eligible call volume, containment, transfer rates, current answering-service costs, loaded staff costs, recruiting expenses, open requisitions, and staff callback time. Revenue should be counted only when improved access produces incremental completed visits or net collections. Abandonment, repeat calls, complaints, and escalation accuracy should remain visible so an apparent cost reduction does not come at the expense of patient access.
Is per-call AI cheaper than hiring medical front-desk staff?
Per-call AI can cost less than adding front-desk capacity at high volume, but the comparison must use contained calls and fully loaded labor rather than salary alone. The national healthcare receptionist wage benchmark is $19 per hour, while benefits and other employer costs raise the total. AI savings should include only hires, overtime, outside coverage, or staff hours that can genuinely be avoided or redeployed. BLS receptionist compensation data
What should a 30-location medical group use instead of a per-call answering service?
A 30-location athenaOne group should evaluate an integrated platform that completes routine requests and writes results into athenaOne, rather than comparing answering rates alone. A conventional answering service can still make sense for narrowly defined human coverage, but message taking leaves callback and data-entry work with the practice. Normalize both options by cost per completed request, then test how pricing changes during peak volume and as locations are added.
How much is an AI medical receptionist worth?
An AI medical receptionist is worth no more than the measurable operating value it creates. That value can include answering-service spend removed, hires or overtime avoided, staff hours reassigned to needed work, and net contribution from incremental completed visits. Gross scheduled revenue, every answered call, and every released staff minute should not be treated as realized return. The defensible figure is monthly benefit actually captured minus recurring AI and oversight costs.
Should a specialty group compare voice AI with an offshore medical call center?
Yes, but the comparison should include workflow completion, EHR writeback, management overhead, coverage, and cost per completed request, not hourly labor rates alone. An offshore team can provide human flexibility for complex or changing work, while voice AI can offer more consistent economics for high-volume, rules-based requests. The practical choice depends on how much work must be completed inside athenaOne and how many exceptions require human judgment. Voice AI versus an offshore medical call center
References
- Pretty Good AI pricing
- Pretty Good AI medical AI receptionist pricing guide
- Talkie.ai pricing
- OhMD pricing
- CallMyDoc ROI and pricing model
- notifyMD medical answering-service pricing
- U.S. Bureau of Labor Statistics receptionist wages
- U.S. Bureau of Labor Statistics employer compensation costs
- SHRM 2025 recruiting benchmarks