The short answer
An AI receptionist answers the calls nobody is free to take, books or routes the request, and hands the rest to a named person.
It matters because 30 to 48 percent of callers never reach a person. Published prices run $29 to $349 a month for software and $300 to $2,100 with human backup.
In this article 9 sections
The five lessons
Read these if you read nothing else.
- Nearly every call is real demand: 92 percent of 25 million calls answered for small businesses were opportunities (Ruby).
- A quarter to a half of callers reach nobody: 30 percent of calls go unanswered (CallRail), and 48 percent of home-services calls are not answered by a person (Invoca).
- The callback rarely works: up to 85 percent of unanswered callers never call back (CallRail).
- Price it from your own missed-call minutes: $29 to $349 a month software-only, $300 to $2,100 human-backed.
- Say it is an assistant and name the human behind it: 89 percent want to know when it is AI (Salesforce).
What the call data says
The largest open call datasets, with sample sizes.
- 48%of home-services calls are not answered by a person, 70 million tracked calls, Invoca 2026
- 30%of calls to businesses go unanswered, CallRail 2025
- 79%of consumers will switch to the business that responds fastest, Invoca 2026
- Up to 85%of unanswered callers do not call back, CallRail 2025
See the samples behind these numbers
| Source | Sample | Finding |
|---|---|---|
| Invoca, 2026 | 70 million tracked calls | 48% of home-services calls are not answered by a person. 79% of consumers will switch to the business that responds fastest. 60% waited over an hour for a form response. |
| CallRail, 2025 | 220,000+ businesses; industry rates from 1.1 million leads | 30% of calls to businesses go unanswered across its customer base. Missed-call rates by industry: healthcare 32%, legal 28%, home services 14%, real estate 9%. Up to 85% of unanswered callers do not call back. |
| Ruby, 2023 call trends | 25 million calls answered for small businesses | 92% of calls are opportunities to win, retain or lose business; only 8% are spam or robocalls. Volume peaks 9 AM to 2 PM and slows between 8 PM and 6 AM. |
| Pew Research Center, 2020 | US adults, national survey | Eight in ten do not answer calls from unknown numbers; 67% check voicemail instead. Your callback is an unknown number. |
| Zendesk CX Trends, 2025 | Consumers and CX leaders, global | 74% of consumers expect customer service to be available 24/7 because AI exists. |
| Salesforce, 2023 and 2025 | Customers and business buyers, global | 77% expect to interact with someone immediately; 89% want to know whether they are talking to AI; 46% of business buyers would work with an AI agent for faster service. |
What an AI receptionist actually does
An AI receptionist answers when a person cannot. It collects a name, number and job, checks a calendar, books or routes the request, and writes the CRM record. It does not diagnose a furnace, quote a roof or close a sale, and the ones that try are the ones owners regret.
Good is measurable. In the 2026 lead response benchmark, Megawebvision-run accounts acknowledged every lead in a median of 10 seconds, texted back every missed call within 2.3 minutes, and replied to 92 percent of message leads inside five minutes, against 0.1 percent industry-wide.
The fit is a few people on the phones, a real after-hours gap, and jobs worth enough that one recovered booking a month covers the subscription. The after-hours share ran from 9 to 53 percent by business, so count yours first. The AI receptionist statistics list every number with its source.
What it costs at published prices
Published list prices on October 4, 2026: software-only AI, AI inside your CRM, and services with a human behind the AI.
| Vendor and type | Plans | Included | Overage |
|---|---|---|---|
| Dialzara, software-only | $29, $99, $199, $349 a month | 60, 220, 500, 1,000 minutes | $0.48 to $0.40 a minute |
| Goodcall, software-only | $79, $129, $249 a month per agent | Unlimited minutes; 100, 250, 500 unique customers | $0.50 per customer above the cap |
| HighLevel, inside the CRM | Starter $97 a month, plus AI Employee $50 to $97 a month per account | CRM, calendar and contact record in one system | Platform usage |
| Smith.ai, human-backed | $300, $810, $2,100 a month | 30, 90, 300 calls | $11.50 to $8.50 a call; $1.50 a call to book |
| Ruby, human-backed | $250 to $720 a month | 50 to 200 minutes; chat bundles from $115 for 10 chats | Per plan |
- Software-only costs a fraction of human-backed service and handles routine capture well. It struggles with callers who refuse a machine and the one call in fifty that needs judgment.
- Human-backed costs five to ten times more per call. It fits legal intake, professional practices and any business where a mishandled call loses a client.
- Inside the CRM is cheapest to run well when you already use the platform.
A worked example: a landscaping company in Austin
An illustrative example with made-up inputs. A landscaping company with eight crews gets most calls between 7 and 9 AM and after 5 PM, when the owner is on a job. The office is staffed 9 to 4, and a third of calls hit voicemail.
She picks a software-only plan for 500 minutes at $199 a month and connects the crews' Google Calendar. The script says it is an assistant, asks for the service and address, offers two estimate slots, and texts her anything urgent. Setup takes an afternoon and ten test calls.
In month one the system books 22 estimate visits from calls that would have gone to voicemail. Four become jobs, and the first paid for the subscription.
The number that made the decision was the voicemail count in her call log for the prior 30 days, not the vendor's demo. Pull yours before you shop. The lead response time statistics show what a slow answer costs.
Where an AI receptionist fails, and the fix for each
Each failure is a decision made before go-live. A better model fixes none of them.
No human backstop
- What it looks like
- The system escalates and nobody picks up
- The fix, decided before go-live
- A named person by hour, and a ten-minute rule if they do not answer
A calendar that is not real
- What it looks like
- Books slots crews do not honor, or never books because the calendar is private
- The fix, decided before go-live
- One calendar the crews live by, shared to the system, checked with ten test bookings
Asked to sell
- What it looks like
- Scripts that quote prices or push upsells
- The fix, decided before go-live
- Capture details, book the visit, stop. A person quotes prices
Nobody reads the transcripts
- What it looks like
- The same mistake repeats a hundred times
- The fix, decided before go-live
- Read every call the next morning for two weeks, then sample a third
CRM record incomplete
- What it looks like
- A name and number with no source, service line or urgency
- The fix, decided before go-live
- The fields are specified before setup, and every record carries them
Five checks to run before you buy
Run these in order, under an hour each.
- Count the gap. From your call log, count missed calls and voicemails by hour and weekday for 30 days. CallRail gives the comparison: 14 percent missed in home services, 28 percent all-business. If your count is small, you need a ring group, not a receptionist.
- Measure your after-hours share from 90 days of CRM timestamps. The benchmark range is 9 to 53 percent by business. Under 15 percent a mobile ring group covers it. Over a third, a receptionist earns its keep.
- Confirm the calendar and CRM are real. Every bookable slot must live in one calendar the crews honor, and every lead must land in the CRM with a source.
- Write the handoff rule: what the AI books alone, what it texts to a person, who that is by hour, and what happens if they do not answer in ten minutes.
- Buy by minutes, not features. Multiply missed calls by average call length, add 30 percent, and buy that tier.
Software-only, in the CRM, or human-backed
Three questions usually decide it.
Service trade, after-hours and overflow calls, jobs booked from a calendar
- Best fit
- Software-only, or inside the CRM
- Why
- Routine capture at the lowest cost; the calendar and record are the product
Already running HighLevel or a similar CRM with a calendar
- Best fit
- Inside the CRM
- Why
- No integration to build; the handoff is a workflow; one record per lead
Legal, medical or financial intake, or any call where a mishandled caller is a lost client
- Best fit
- Human-backed
- Why
- Judgment on the call, at five to ten times the per-call cost
Callers who refuse to talk to a machine are a known share of your base
- Best fit
- Human-backed, or software-only with instant transfer
- Why
- Transparency plus a fast path to a person keeps the caller
- A practical path: start software-only or inside the CRM for after-hours and overflow, read transcripts for two weeks, and go human-backed only if they show calls the AI should not have taken.
What this means for your next decision
A quarter to a half of inbound phone demand is at risk, the callback rarely works, and callers reward the fastest responder. An AI receptionist is the cheapest fix.
Price it from your own missed-call count, give it a real calendar and a named human backstop, say what it is, and read transcripts for two weeks. The diagnostic shows whether response is your constraint. AI implementation cost covers the full system.
Questions leaders ask
How much does an AI receptionist cost for a small business?
At published prices on October 4, 2026, software-only plans run from $29 a month for 60 minutes to $349 for 1,000 minutes, and human-backed services from $300 a month for 30 calls to $2,100 for 300 calls. AI inside HighLevel adds $50 to $97 a month to the platform plan.
How many calls does a small business actually miss?
CallRail puts 30 percent of calls unanswered across its customer base, and its 1.1 million-lead benchmark runs from 14 percent in home services to 32 percent in healthcare. Invoca's 70 million calls found 48 percent of home-services calls are not answered by a person.
Will callers hang up on an AI receptionist?
Some will, mostly on complex or emotional calls, but most callers who reach voicemail hang up anyway and up to 85 percent never call back. Compare the AI against voicemail, and tell callers it is an assistant.
Can an AI receptionist book appointments?
Yes, when it is connected to one calendar your staff actually honor, and it writes the contact into the CRM. The common failure is a private or stale calendar, so it books into slots that do not exist.
Should an AI receptionist replace the front desk?
For most small businesses, no. It covers evenings, weekends, lunch and the second line, while a person who knows the business converts better during staffed hours.
SOURCES
- Invoca (2026), citing its Lead Conversion Benchmarks Report (70 million calls) and Home Services Buyer Experience Report: 48 percent of callers never speak with a person; 60 percent waited more than an hour for a form response; 79 percent switch to the fastest responder
- CallRail, Small Business Marketing Benchmark Report (January 2025): 1.1 million leads; missed call rates by industry; up to 85 percent of unanswered callers will not call back
- CallRail (September 2025): 30 percent of calls to businesses go unanswered (July 2025 release)
- Ruby, Call Trends Report: 25 million answered calls; 92 percent are opportunities; 8 percent spam; volume peaks 9 AM to 2 PM Pacific
- Pew Research Center (2020): eight in ten Americans do not generally answer calls from unknown numbers; 67 percent check voicemail instead
- Zendesk CX Trends 2025: 74 percent of consumers expect 24/7 service because of AI
- Salesforce (2023): 77 percent expect to interact with someone immediately; 89 percent want to know if they are talking to AI
- Salesforce, State of the Connected Customer: 46 percent of business buyers would work with an AI agent for faster service; 72 percent say it is important to know if they are communicating with an AI agent
- Nielsen Norman Group, The User Experience of Chatbots: bots less helpful than humans, valued for speed, transparency about the bot welcomed
- Dialzara pricing: $29 to $349 a month, overage $0.40 to $0.48 a minute
- Goodcall pricing: $79, $129 and $249 a month per agent
- HighLevel pricing: Starter $97 a month, AI Employee $50 to $97 a month per sub-account
- Smith.ai pricing: $300 a month for 30 calls to $2,100 for 300, overage $8.50 to $11.50 a call
- Ruby pricing: $250 a month for 50 minutes to $720 for 200 minutes; chat bundles from $115
- Megawebvision, Lead Response Time Benchmark 2026: run-account reply speed and missed-call text-back
Where’s the money leaking?
Pick one. Be honest.
Cite this article
Grigorchuk, T. (2026, October 8). AI Receptionist for Small Business: 70 Million Calls Say Why, Published Prices Say What It Costs. Megawebvision. https://megawebvision.com/insights/ai-receptionist-for-small-business