Why comparing quotes is deliberately hard
One provider quotes a per-minute rate. Another quotes a per-call rate. A third quotes a monthly plan with an included allowance. A fourth quotes a per-receptionist-minute rate that rounds every call up to the next full minute. None of these are directly comparable, and that is not an accident.
The only number that matters is your total monthly cost at your actual call volume, including the calls you did not want. Until a provider gives you that, you do not have a quote, you have a rate card.
The good news is that translating between models is straightforward arithmetic once you know your own numbers. The bad news is that most owners do not know their own numbers, which is exactly what the pricing model relies on.
So before you ring anyone: pull your call log for a typical month. How many inbound calls? How long was the average one? How many were wrong numbers, robocalls, or suppliers? That last figure is where the money hides.
The four pricing models, translated
Per-minute is the most common and the most misleading. The rate looks small, but billing usually rounds up to the next minute or the next six seconds, and the clock often starts before a human picks up. Ask specifically: when does billing start, what is the rounding increment, and is hold time billed?
Per-call is easier to reason about and aligns better with value, since a call is roughly a lead. The trap is the definition. Ask whether a hang-up counts, whether a wrong number counts, and whether a caller who rings back twice counts as one or two.
Monthly plans with an included allowance are the easiest to budget and the easiest to overrun. The critical question is the overage rate, because it is frequently much higher than the effective rate inside the plan. A plan that is comfortable in January can be painful in your busy season.
Flat monthly pricing, usually with a one-off setup fee, is what most done-for-you AI providers use. It removes the volume anxiety entirely, which matters more than people expect, because per-minute pricing quietly discourages you from sending it more calls.
The fees that are not in the headline number
Setup and onboarding. Sometimes free, sometimes several hundred dollars, and occasionally buried as a first-month surcharge. Ask whether it is refundable if the thing does not work.
Script or knowledge changes. Some human services charge per script revision. If your business changes seasonally, this adds up, and worse, it discourages you from keeping the script accurate.
After-hours, weekend, and holiday premiums. These are common with human services and they apply to exactly the hours you bought the service for. An after-hours premium on an after-hours answering service deserves a straight answer about what the effective rate really is.
Integrations, extra numbers, call recording, transcription, and reporting are all sometimes separate line items. And check the contract length and the notice period, because a twelve-month lock-in changes the risk of trying something entirely.
Human services versus AI, on cost alone
Human answering services price on labour, so their cost scales almost linearly with your call volume. That is fair, and it means a busy month is an expensive month. Their strength is judgement and genuine empathy on a difficult call.
Dedicated virtual receptionists cost substantially more per hour because someone is actually learning your business. For firms where every call is high value, this can be the right answer and the cost is not the point.
AI answering costs are dominated by the build rather than the volume, so the marginal cost of one more call is close to nothing. That inverts the incentive: you want it to take more calls, not fewer. The weakness is the judgement gap I discuss in what is an AI receptionist.
If what you are really deciding is whether to put a person on the front desk instead, the full comparison including the costs nobody quotes is in AI receptionist versus a front-desk hire.
Work out your own number first
Take your monthly inbound call count and multiply by your average call length. That is your billable minutes under a per-minute model, before rounding, which typically adds ten to twenty percent.
Now subtract nothing. That is the point. Under per-minute and per-call models you pay for the supplier confirming a delivery and the robocall about your car warranty, unless the contract explicitly says otherwise.
Then work out what a booked job is worth to you on average, and how many calls a month currently go unanswered. If recovering even a fraction of those covers the cost, the pricing model matters far less than simply having something answer.
That is genuinely the whole calculation. A trade with a four-figure average job and six missed calls a month is in a completely different decision than a salon with a sixty-dollar service, and no provider's pricing page will tell you which one you are.
What to ask before you sign anything
What will my total bill be at my actual volume, including wrong numbers and hang-ups? Ask for it in writing, as a number, not a rate.
What is the overage rate, and what happened to your busiest customers last December? A provider who cannot answer the second question is not paying attention to their own product.
What is the contract length, the notice period, and what do I keep if I leave? Call recordings, transcripts, and the contact records generated by those calls should be yours.
And can I hear it work before I pay anything? For an AI service that is a reasonable ask. Ask me and I will put you on a live one. For a human service, ask to hear a recording of a real call handled for a business like yours.