Short answer
An off-the-shelf AI receptionist costs roughly US$50 to US$500 a month for a small business, compared with about US$250 to US$2,100 a month for a live answering service and a median wage of US$38,010 a year for a full-time human receptionist in the United States. The raw voice technology underneath is cheap: telephony plus a voice AI layer runs around eight to nine US cents a minute at published pay-as-you-go rates. What you are really paying for is setup, integration with your calendar and CRM, and the rules that decide what the receptionist may say and do. Below: sourced prices for each option, a worked monthly comparison, what an AI receptionist can and cannot do, recording consent rules, and how to measure it.
The short answer: four ways to answer the phone
There are four ways to make sure calls get answered: hire a receptionist, pay a live answering service, subscribe to an AI receptionist, or have one built on your own phone line and systems. A human costs the most and handles the widest range of situations; an AI receptionist costs the least per call and never sleeps, but it only does what it has been set up and allowed to do.
The comparison matters because unanswered calls are common. In a 2016 study, 411 Locals monitored the calls of 85 small businesses in 58 industries for 30 days: only 37.8% of calls were answered by a person, another 37.8% went to voicemail and 24.3% got no response at all. Every one of those calls is a potential customer who may simply call the next business on the list.
| Option | How it is priced | Published price range (USD) | Best for |
|---|---|---|---|
| In-house receptionist | Hourly wage plus employer costs | US median US$18.27/hour, US$38,010/year (May 2025) | Busy front desks with in-person visitors |
| Live answering service | Per minute or per call, monthly bundles | Ruby US$250–US$1,725/month; Smith.ai US$300–US$2,100/month | Calls that need human judgment and empathy |
| AI receptionist (SaaS) | Monthly plan by calls, minutes or callers | Rosie from US$49/month; Goodcall from US$79/month; Smith.ai AI from US$0 to US$500/month | Routine calls, after-hours coverage, simple booking |
| Custom AI receptionist | Setup fee plus monthly operation; usage at cost | Varies by integrations; see ZeniTech prices below | Booking and intake inside your own calendar, CRM and rules |
Wages: US Bureau of Labor Statistics, May 2025. Service prices from each vendor’s pricing page as viewed on September 28, 2026, before taxes. The 411 Locals figures come from a small 2016 sample and should be read as an indication, not a national rate.
What a human receptionist costs
A full-time receptionist in the United States earns a median of US$18.27 an hour, or US$38,010 a year, according to the Bureau of Labor Statistics (May 2025); the lowest 10% earned under US$13.83 and the highest 10% over US$24.01 an hour. In Canada, the federal Job Bank reports a median of C$21.00 an hour for receptionists nationally and C$20.00 in Québec (2023–2024 data). Payroll taxes, benefits, paid leave and training come on top of the wage.
The bigger constraint is coverage, not the hourly rate. One full-time person covers about 40 of the 168 hours in a week. Covering evenings and weekends with people means shifts, overtime or a second service, which is why many businesses combine a receptionist during office hours with an answering service or an AI receptionist for overflow and after-hours calls.
Live answering services: sourced prices
Live answering services put a professional receptionist on your line, billed by the minute or by the call. They are the right choice when callers are upset, the conversation is unpredictable, or a mistake would be costly. Their weakness is price per call, which rises quickly with volume.
Ruby sells receptionist minutes in monthly bundles, from US$250 for 50 minutes to US$1,725 for 500 minutes, with no setup fee; its pricing page does not publish an overage rate. Smith.ai prices its human receptionists per call: US$300 a month for 30 calls, US$810 for 90 and US$2,100 for 300, with overage of US$11.50, US$10.50 and US$8.50 per extra call respectively, month to month with a 10% discount for annual prepayment. AnswerConnect does not publish prices on its site; you request a quote.
| Service and plan | Monthly price (USD) | Included | Effective cost |
|---|---|---|---|
| Ruby Starter | US$250 | 50 minutes | US$5.00 per minute |
| Ruby Popular | US$720 | 200 minutes | US$3.60 per minute |
| Ruby Growing Business | US$1,725 | 500 minutes | US$3.45 per minute |
| Smith.ai Starter (human) | US$300 | 30 calls | US$10.00 per call; US$11.50 over |
| Smith.ai Basic (human) | US$810 | 90 calls | US$9.00 per call; US$10.50 over |
| Smith.ai Pro (human) | US$2,100 | 300 calls | US$7.00 per call; US$8.50 over |
Effective per-minute and per-call costs are ZeniTech’s arithmetic from the published bundle prices (Ruby and Smith.ai pricing pages, viewed September 28, 2026).
AI receptionist SaaS: what the plans cost
AI receptionist subscriptions are an order of magnitude cheaper per call than human services, and they price in three different ways: per call, per minute or per unique caller. Compare them on your own call volume and average call length, not on the headline price.
Smith.ai’s AI receptionist has a free plan with 25 calls a month (US$3.00 per extra call), a Pro plan at US$150 a month starting at 75 calls, and an Enterprise plan at US$500 starting at 300 calls, with overage between about US$1.67 and US$2.17 per call depending on the tier. Goodcall charges per agent and per unique customer rather than per minute: US$79, US$129 or US$249 a month for 100, 250 or 500 unique customers, then US$0.50 per extra customer, with 15% off for annual billing. Rosie sells minutes: US$49 a month for 250 minutes, US$149 for 1,000 and US$299 for 2,000, with a 7-day trial.
| AI service | Pricing unit | Entry plans (USD/month) | Watch for |
|---|---|---|---|
| Smith.ai AI Receptionist | Per call | US$0 (25 calls), US$150 (from 75 calls), US$500 (from 300 calls) | Overage US$1.67–US$3.00 per call |
| Goodcall | Per unique customer, per agent | US$79 (100), US$129 (250), US$249 (500 customers) | US$0.50 per extra customer; one agent per plan |
| Rosie | Per minute bundle | US$49 (250 min), US$149 (1,000 min), US$299 (2,000 min) | Overage rate not published on the pricing page |
Prices from the Smith.ai AI receptionist, Goodcall and Rosie pricing pages as viewed on September 28, 2026, monthly billing, before taxes.
What the voice AI itself costs per minute
Underneath every AI receptionist there are three metered components: the phone line, speech (listening and speaking), and the language model that decides what to say. At published pay-as-you-go rates, the first two together cost roughly eight to nine US cents a minute; the language model usually adds little for short calls.
Twilio lists inbound calls to a US local number at US$0.0085 a minute (US$0.022 on toll-free), a local number at US$1.15 a month, call recording at US$0.0025 a minute and its ConversationRelay voice AI connector at US$0.07 a minute. Deepgram lists its Voice Agent API at US$0.075 a minute on the standard tier, and streaming speech recognition (Nova-3) from US$0.0048 a minute at a promotional rate. ElevenLabs lists additional agent minutes at US$0.08. The table below prices an illustrative month; it is an example calculation, not a client measurement.
| Illustrative month: 150 calls, 3 minutes each (450 minutes) | Calculation | Approx. cost (USD) |
|---|---|---|
| Twilio local line + ConversationRelay | 450 × (US$0.0085 + US$0.07) + US$1.15 number | about US$36 |
| Twilio local line + Deepgram Voice Agent (standard) | 450 × (US$0.0085 + US$0.075) + US$1.15 | about US$39 |
| Twilio local line + ElevenLabs agent minutes | 450 × (US$0.0085 + US$0.08) + US$1.15 | about US$41 |
| Smith.ai AI Pro (150 calls) | US$150 for 75 calls + 75 extra calls at about US$2 | about US$300 |
| Smith.ai Basic, human (150 calls) | US$810 for 90 calls + 60 × US$10.50 | US$1,440 |
| Ruby Growing Business (450 minutes) | Bundle of 500 minutes | US$1,725 |
Raw usage rates from the Twilio (US voice), Deepgram and ElevenLabs pricing pages as viewed on September 28, 2026. They exclude language-model tokens, recording, SMS confirmations, engineering and the work of connecting a calendar or CRM, which is where most of the real cost of a custom build sits.
What an AI receptionist can and can’t do
A well-configured AI receptionist handles the predictable majority of calls very well: it answers instantly, around the clock, in more than one language, and it never forgets to ask for the phone number. It is weak on anything that needs judgment, empathy or authority it was not given.
- Can: answer questions about hours, location, services, prices and policies, from content you approve.
- Can: qualify a caller with a few set questions (type of job, urgency, location, insurance or file number) and write a clean summary.
- Can: book, move or confirm appointments directly in a connected calendar, and send a text confirmation.
- Can: route urgent calls to a person, take a message, and handle overflow when your line is busy.
- Can’t: give medical, legal or financial advice, and it must be configured to refuse and escalate instead.
- Can’t: reliably calm an angry caller, negotiate, or make exceptions to your policies. Hand these to a person.
- Can’t: know anything you did not give it. Wrong or outdated source content produces wrong answers, delivered confidently.
- Can’t: replace the in-person side of a front desk: greeting visitors, handling payments at the counter, paperwork.
Call recording consent and AI voice rules
Most AI receptionists record and transcribe calls, so consent rules apply. This is a high-level overview, not legal advice; check the rules for each place your callers are in.
In the United States, federal law allows recording with the consent of one party to the call, but a minority of states, about a dozen including California, Florida, Illinois, Maryland, Massachusetts, Montana, New Hampshire, Pennsylvania and Washington, require the consent of all parties. The Digital Media Law Project advises that when a call involves people in more than one state, the safe course is to get everyone’s consent. Separately, the FCC ruled in February 2024 that AI-generated voices are “artificial” under the Telephone Consumer Protection Act, so outbound calls made with an AI voice need the called person’s prior express consent. That matters if your receptionist also calls leads back.
In Canada, the Criminal Code allows interception with the consent of one party, but businesses are also bound by privacy law. The Office of the Privacy Commissioner of Canada says an organization recording customer calls must tell the caller, state the purpose and obtain consent, and offer an alternative if the caller objects. Québec’s Law 25 and, for European callers, the GDPR add their own obligations on personal information.
- Announce at the start of every call that the caller is speaking with an AI assistant and that the call is recorded, and why.
- Offer a way to reach a person, or to leave a message without being recorded.
- Decide how long recordings and transcripts are kept, and where they are stored.
- Get prior consent before any AI-voiced outbound call, text campaign or follow-up sequence.
Lead capture and booking: where the money is
An AI receptionist pays for itself through the calls it turns into appointments, not through the salary it saves. Set it up to finish the job on the first call whenever possible, and to leave your team with a complete, usable record when it cannot.
- Book directly into the real calendar, with the service length and buffers you choose, instead of promising a callback.
- Capture the caller’s name, number, reason for calling and consent to be contacted, then create or update the contact in your CRM with the transcript attached.
- Send an immediate text confirmation and a reminder before the appointment to reduce no-shows.
- Alert a person right away for high-value or urgent calls: an emergency repair, a new legal matter, a patient in pain.
- Recover missed calls: if a caller hangs up before speaking, a text asking how you can help often restarts the conversation. Check consent rules first.
- Review a sample of transcripts weekly; unanswered questions tell you exactly which content to add.
How to measure an AI receptionist
Measure outcomes against a baseline. Before launch, pull a month of call logs from your phone provider: how many calls came in, how many were answered, how many went to voicemail, and how many turned into bookings. Compare after 30 and 90 days.
| Metric | What it tells you | How to measure |
|---|---|---|
| Answer rate | Whether calls are no longer lost | Answered calls divided by inbound calls, from phone provider logs |
| Bookings from calls | Direct revenue contribution | Appointments created by the receptionist in the calendar or CRM |
| Lead capture rate | Whether callers leave usable details | Calls with a name, number and reason recorded, divided by calls handled |
| Transfer and escalation rate | Where the AI’s knowledge or authority stops | Calls handed to a person, with the reason, read weekly |
| Answer accuracy | Risk of wrong information | Weekly transcript sample checked against your approved content |
| Cost per handled call | Whether it beats the alternatives | Total monthly cost divided by calls handled end to end |
| Complaints and hang-ups | Caller experience | Early hang-ups, negative feedback, and requests to “speak to a human” |
What ZeniTech charges
ZeniTech is a web, app, automation and AI agency based in Québec City, working remotely with clients across North America and Europe on Eastern Time, in English or French. We build AI receptionists that answer calls, texts and website messages, book into your calendar, and send each lead to your CRM with a summary. Our agents are powered by Orvel AI, ZeniTech’s agent layer operated in Québec, and we have ready-made versions for dental clinics, law firms and auto repair shops.
Every project starts with a written scope and a fixed price per phase. The code repository is in the client’s name from the first commit. The monthly plan includes reasonable usage; exceptional volumes are agreed in advance. For low call volumes and simple needs, an off-the-shelf AI receptionist is often cheaper, and we will say so.
| Service | Price (CAD, before taxes) | Approx. USD |
|---|---|---|
| AI receptionist or agent setup (powered by Orvel AI) | $3,000 to $7,500 | approx. $2,120 to $5,290 |
| Operation and improvement | $750 to $1,500/month | approx. $530 to $1,060/month |
| Simple automation (e.g. missed call to text and CRM) | $750 | approx. $530 |
| Complex workflow (several systems, branching, error handling) | $2,500 | approx. $1,765 |
| Monitoring of automations | $250/month | approx. $176/month |
| Work outside the agreed scope | $125/hour | approx. $88/hour |
Conversions use the Bank of Canada rate of 1.4168 CAD per USD (September 28, 2026) and are rounded. You are invoiced in CAD.
Frequently asked questions
How much does an AI receptionist cost for a small business?
Off-the-shelf AI receptionists start around US$49 to US$150 a month: Rosie is US$49 for 250 minutes, Goodcall US$79 for 100 unique callers, and Smith.ai’s AI Pro plan US$150 from 75 calls. Higher volumes run a few hundred dollars a month. A custom AI receptionist connected to your calendar and CRM costs more upfront; at ZeniTech setup is CA$3,000 to CA$7,500, then CA$750 to CA$1,500 a month.
What is the difference between a virtual receptionist and an AI receptionist?
A virtual receptionist is usually a live person working remotely for an answering service such as Ruby or Smith.ai, billed per minute or per call, from about US$3.45 a minute to US$11.50 per extra call on published plans. An AI receptionist is software that answers with a synthetic voice, costs a small fraction of that per call and works around the clock, but only handles what it has been set up and permitted to do.
Is an AI receptionist cheaper than hiring a receptionist?
In cash, almost always. The US median receptionist wage is US$38,010 a year before employer costs (BLS, May 2025), while AI plans cost hundreds, not thousands, of dollars a month. But they do different jobs: a person greets visitors and handles difficult conversations. Many businesses keep their receptionist and use AI for overflow, lunch breaks, evenings and weekends.
How much does voice AI cost per minute?
At published pay-as-you-go rates, roughly eight to nine US cents a minute for the phone line plus the voice AI layer. Twilio lists inbound local calls at US$0.0085 a minute and ConversationRelay at US$0.07; Deepgram’s Voice Agent API is US$0.075 a minute and ElevenLabs lists extra agent minutes at US$0.08. Integration work, not minutes, is where most of the cost sits.
Is it legal to record calls with an AI receptionist?
Yes, with the right notice. US federal law requires one party’s consent, but about a dozen states, including California, Florida and Pennsylvania, require all parties’ consent. In Canada, the Privacy Commissioner expects businesses to tell callers they are recorded, why, and to obtain consent. Announcing the AI and the recording at the start of every call covers most situations.
Can an AI receptionist book appointments?
Yes, if it is connected to your real calendar or booking system. It can check availability, book, move or cancel appointments within the rules you set, and send a text confirmation and reminders. Without that connection it can only take a message, which loses much of the value. Give it permission to book, not to cancel or refund, unless you have tested those flows.
Will callers hang up on an AI receptionist?
Some will, especially if the AI hides what it is or traps them in loops. Say clearly at the start that it is an AI assistant, keep answers short, and offer a person or a message at any time. Then measure early hang-ups and “speak to a human” requests weekly; if they stay high, the script or the content needs work.
What should an AI receptionist never do?
It should never give medical, legal or financial advice, quote prices or policies that are not in its approved content, promise exceptions, or make outbound AI-voiced calls without prior consent. It should also never pretend to be a person. Configure it to refuse, take a message and escalate in these cases, and log every call for review.
Related
Free consultation
Get a fixed price for your project in 30 minutes.
We look at what the project has to achieve, tell you what it costs, and tell you when you do not need it.
Book a call →+1 581-748-7017Sources
- US Bureau of Labor Statistics: Receptionists, Occupational Outlook Handbook (May 2025 data)
- Government of Canada Job Bank: Receptionist wages, NOC 14101 (2023–2024 data)
- 411 Locals: Small business owners don’t answer 62% of phone calls (2016)
- Ruby: Pricing (viewed September 2026)
- Smith.ai: Receptionist pricing (viewed September 2026)
- Smith.ai: AI Receptionist pricing (viewed September 2026)
- AnswerConnect: Pricing (quote on request, viewed September 2026)
- Goodcall: Pricing (viewed September 2026)
- Rosie: Pricing (viewed September 2026)
- Twilio: Voice pricing, United States (viewed September 2026)
- Deepgram: Pricing (viewed September 2026)
- ElevenLabs: API pricing (viewed September 2026)
- Digital Media Law Project: Recording phone calls and conversations
- FCC: FCC makes AI-generated voices in robocalls illegal (February 8, 2024)
- Justice Laws Website: Criminal Code, section 184
- Office of the Privacy Commissioner of Canada: Recording of customer telephone calls
- Bank of Canada: Daily exchange rates (September 28, 2026)