AI receptionist pricing, explained: per-minute, per-call, per-conversation — and flat

Every AI receptionist is priced on a meter. The only questions are what the meter counts, and what happens when you pass the line. Here are the four models in the market today, with the real prices attached, so you can do the math for your shop before you sign anything.

SavedRing · Pricing ·

  • Graded 19/21 · 20/21, Grade A, on our published 21-point scorecard
  • Demo line: (571) 702-6699
  • Founding $199/mo flat (first 15 shops) · Standard $249/mo

If you are shopping for an AI receptionist, the headline prices look simple: $49 a month, $149 a month, $150 a month. They are not comparable numbers, because each one buys a different unit. One buys minutes. One buys calls. One buys conversations. Until you know which unit a vendor is selling, you cannot know what a busy month costs — and in this market, the busy month is where the pricing actually lives.

Model 1: Per minute

The oldest model, inherited from human answering services. You buy a pool of minutes, and every answered call draws the pool down.

Rosie is the clearest published example. Its plans are $49 a month for 250 minutes, $149 for 1,000 minutes, and $299 for 2,000 minutes. Three things about a minute pool are easy to miss on a pricing page. First, calls are billed in 60-second increments, rounded up, so a 61-second call bills as two minutes. Second, every answered call draws the pool down — the real customer, the repeat caller, the wrong number that stays on the line. Third, what happens when the pool runs out is a policy, not a law of nature, and vendors choose very different policies. Rosie's is an automatic upgrade to the next tier, which stays in place until you downgrade it yourself. (That mechanism gets its own article on this blog.)

Human services run the same meter at higher rates. Ruby bills per minute — $250 a month for 50 minutes up to $1,725 for 500 — and the recurring theme across its public reviews is that junk calls eat the meter: spam and robocalls billed as minutes, at the same rate as a customer with a burst pipe. Ruby's hold-time and rounding billing practices were the subject of a federal class action that settled for up to $12 million. The meter itself was the complaint.

What to check: the rounding rule, what counts against the pool, and the overage policy — in that order.

Model 2: Per call

Here the unit is the call itself, short or long. Smith.ai's AI plan is $150 for 75 calls. That is $2.00 per included call, and every call after the 75 costs $2.50. A shop that takes 100 calls in a month pays $212.50, not $150.

Per-call pricing has one virtue: a long, complicated call costs the same as a short one. Its hazard is the add-on menu that tends to grow around the base plan, and per-call overage that stacks up exactly in your busiest months — Smith.ai's human receptionist overage runs $8.50–$11.50 per call.

What to check: the per-call overage rate, and a written list of everything that is not included in the base plan.

Model 3: Per conversation

Newer, and easy to confuse with per-call. A conversation is a handled exchange with one customer, and the meter runs per unit, all month.

If your shop runs on Jobber, its AI receptionist is an add-on: $29 a month, 30 conversations included, then $0.79 per conversation after that — on top of the Jobber plan itself. GorillaDesk's AI runs $1.00 per informational conversation and $1.75 per action conversation. The per-unit charge sounds small, and it is. It is also permanent arithmetic that grows with your business: the better your marketing works, the more you pay for answering it.

A variant counts people instead of conversations. Goodcall charges per unique caller: $79 a month for 100 unique callers, $129 for 250, $249 for 500, then $0.50 per additional caller — with unlimited minutes per caller.

What to check: what exactly counts as one unit, whether your existing software subscription is part of the real price, and what happens at the cap — a charge, or a pause.

Model 4: Flat

One price, no meter running per unit. This is how SavedRing is priced, so here is our own math in the same open form.

SavedRing is front-desk software for home-service shops, with an AI receptionist — Donna — built in. Every handled conversation becomes a job card in your portal: caller, problem, address, urgency, and the visit window requested, with the recording and transcript attached.

The price is $199 a month for the first 15 shops (Founding), then $249 a month (Standard). No setup fee. No contract — month-to-month. The free pilot is 14 days at $0, no card required.

The flat price includes 150 handled conversations a month, and "handled" is defined narrowly on purpose: a conversation counts only when Donna actually speaks to a person. Voicemail attempts, spam, robocalls, and wrong numbers never count. Past 150 there is no per-minute or per-conversation charge — we talk with you about the right tier first. Nothing is billed automatically, and the plan never changes on its own. Switching from a competitor? Any shop switching receives a $99 credit toward its first paid month, on any tier.

What to check: what counts toward the allowance, and what happens past it. If the answer to the second question is a charge or an automatic upgrade, it is not flat — it is a meter with a friendly first page.

The one-sentence version

Minutes punish you for long calls, per-call punishes you for short ones, per-conversation charges you for growing, and flat puts the counting risk on the vendor — which is where we think it belongs.

Hear the receptionist before you price anything: call Donna, our live demo line, at (571) 702-6699 and grade her yourself. Or take the free 14-day pilot and watch the job cards land in your own portal.

Prices for Rosie, Ruby, Smith.ai, Jobber, GorillaDesk, and Goodcall are as published by those vendors and in our Nov 15, 2026 counting-unit research (SavedRing's fair-use counting study). Vendor prices change; check the vendor's page for the current figure.