Set the carrier or phone-system forward so that after closing time the main number rings a VoiceGenie number instead of voicemail. Daytime calls still ring your staff. Urgent callers are sent to the on-call line.
AI answers every call. Your people handle the ones that matter.
The AI takes hours, messages, routing and booking on the first ring. Your team keeps the IRS notice, the upset client and the emergency, and that beats a staffed desk or an AI alone.
- Starts with one lane, not the whole line
- Can say it is an AI; a person on request
- Keeps your number, carrier and phone system

Hours, parking, directions
The same answer every time, on the first ring.
Can I speak to Priya?
Look up availability, transfer or take a message.
Message for staff
Name, number and reason read back and sent by text.
Urgent or upset caller
Straight to a person, with the summary.
Why the front desk will run on AI and people together
A front desk gets the same dozen questions all day, then one call that carries real risk. A busy line sends callers to voicemail, and under 3% of them leave a message. A staffed desk cannot cover lunch, 8 pm and four ringing lines. An AI alone cannot calm a client holding a tax notice.
- Picks up on ring one at 8 am, at lunch and at 8 pm
- Gives the same hours, parking and price answer every time
- Takes a message and reads back the name and number
- Calms the client holding the IRS letter
- Quotes the deck job or takes the case on judgment
- Grants the credit, the exception or the fix
- No call reaches voicemail, and a person gets the ones that need judgment
- The preparer hears the notice details first, so the client never repeats them
- Staff stop answering parking questions and spend the day on clients in front of them
A billing call can be a simple question or a dispute. Give the AI calls with one right answer. Give people calls where a wrong answer loses a client.
Start where the alternative is voicemail: after hours, lunch and the busiest hour. Nobody loses a task, and you can count the calls you were losing.
Judge the system by what your person has in front of them on pickup. A summary, a transcript and the caller's mood save the client from starting over.
What does an AI receptionist do, and which calls should it handle?
An AI receptionist answers every call on the first ring, says it is an AI, and finishes the routine ones: hours, directions, prices, message-taking, routing, booking and rescheduling. It hands anything involving judgment, distress or money to a person, with a summary and transcript. Most teams start with after-hours or overflow calls, then widen once accuracy holds.
- US receptionists: median pay $38,010 a year, about 947,500 jobs (BLS, 2025)
- 78% of 1,000 US consumers say they left a business after an unanswered call (CallRail, 2025)
- About 12% of calls to trade shops arrive outside business hours (ServiceTitan sample, 2026)
- AI is $0.14 per minute, billed per second. Human seats start at $29 a month
The front desk is where a first impression is won or lost.
A caller who reaches voicemail rarely leaves a message, and a caller who is transferred badly remembers it. Four numbers show both sides: calls that go unanswered, and calls that are answered poorly.
of 1,000 US consumers say they abandoned a business after an unanswered call. CallRail consumer survey, published Sept 20251
of callers sent to voicemail leave a message. Invoca platform data, home services, article dated May 2024, period not stated2
of calls missed in the highest-rate industries: home services 14%, legal 28%, healthcare 32%. CallRail, 1.1 million leads, 20253
of 3,000 consumers say they have to repeat themselves after being transferred; 35% say often or always. Vendor-commissioned CX survey, US, UK and Germany, 20264
Where the main line shows up in a customer's life with your business
The front desk is the first voice a prospect hears and the one a client calls when something is off. Here is how the AI and your people split the phone at each stage, from the first inquiry to the client who went quiet.
Read down a column for one stage, or across a row to see what the AI and your team each do from first contact to win-back.
Which calls the AI owns, which it starts, and which go straight to a person.
A receptionist's day is a handful of call types repeated. Sort them by what is at stake and how much judgment the answer needs, not by how the caller phrases it. Tag your own top ten call reasons for two weeks before you set this table.
One overflow call, start to finish, from busy line to preparer.
This is the call in the box above, followed step by step. The point is not that the AI answered. The point is what the person had in front of them when they picked up.
Phones ring. The line is busy.
Your phone system rings your team for four rings, then forwards to a VoiceGenie number. Callers never reach voicemail.
AI answers and says what it is
It gives the business name, says it is an AI assistant, says the call is recorded, and offers a person at any time.
It finds the reason and the caller
It asks one open question, then confirms the name and number by reading them back. It does not run a menu.
It checks the record and the schedule
Caller ID is matched to the client record. The AI sees the assigned preparer and that she is free at 9:30.
Warm hand-off to a person
The preparer's phone shows a one-line summary, the transcript, the caller's mood and what the AI already did. The caller repeats nothing.
Notes and follow-up are drafted
AI Assist proposes the disposition, call-back time and notes. The preparer accepts them in one click. The call is recorded and scored.
Start with the calls you are already missing. Widen only when the numbers hold.
Do not flip the whole main line. Pick one lane where the alternative is voicemail, run it for a few weeks with every call reviewed, then add a second lane. Each stage below has a test before the next begins.
After-hours calls: answer, take a message or book, hand urgent ones to on-call
Today these calls reach voicemail, and under 3% of callers leave a message. No one is displaced, the baseline is easy to count, and the worst case is a call you were already losing.
Set up and rehearse
- Pull last month's call log by hour and tag the top ten call reasons
- Write the knowledge base: hours, services, prices, policies, who handles what
- Set the urgent list and the on-call number, and test the transfer from a mobile phone
- Run 30 test calls with staff and friends, including angry and mumbling callers
After-hours live, every call reviewed
- Turn on the after-hours forward only
- Read or listen to every call each morning and fix the knowledge base
- Check messages: number, name and reason correct, sent to the right person
- Count calls that asked for a person and how fast they got one
Add overflow in the busiest hour
- Forward to the AI after four rings during your two busiest hours
- Add booking and rescheduling for one or two appointment types
- Review a sample of 20 calls a day instead of all of them
- Compare answer rate and abandonment with the same weeks last quarter
Add lunch, absences and routing
- Cover lunch and days when someone is out
- Add routing by name and department, with message fallback
- Hold a team review: what did the AI prepare well, what should change
- Decide whether to extend to all daytime overflow
Expand when all of these are true
- Messages: name, number and reason correct on at least 95% of sampled calls
- Every urgent test and live call reached a person in under 60 seconds
- Callers who ask for a person get one every time, with no repeated questions
- After-hours hang-ups before any outcome are below your voicemail baseline
Seven numbers that tell you whether the line is working.
Measure the outcome of the call, not whether the AI picked up. Take a baseline from your own call log first. The benchmarks below are context from published surveys, and the targets are planning ranges, not promises.
Planning ranges are targets to test in a pilot, not promises. Measure your own baseline for two weeks before you change anything.
What an AI receptionist is worth to a mid-sized office.
Use your own numbers in the calculator. The example below is a single office taking about 1,800 calls a month, with a mix of routine questions, messages and new-customer calls.
A worked example. Move the sliders to match yours.
AI at $0.14 a minute with volume credit applied, plus Insights seats for the people who take handed-over calls. Phone line charges are your carrier's and are not included. Illustration, not a quote. Open the full calculator
Where these numbers come from
- Cost per person: BLS median receptionist pay of $18.27 an hour 7, times 1.43 for benefits and taxes from BLS compensation data 8. This is our estimate.
- Missed share of 20% sits between the published industry rates of 14% and 32% 3. Replace it with the share in your own log.
- Calls per person per day (50) and AI finishes (50%) are planning ranges, not published benchmarks. Count both over a normal fortnight.
- Conversion of 30% on recovered calls is near the median trade-shop booking rate 6. Many recovered calls are less qualified than average.
- Outcome value is first-appointment revenue and is a placeholder. Use your average first-visit value.
What the dollar figure leaves out
- Your team stops being interrupted by routine calls and can finish the ones in front of them
- Callers reach someone on the first ring at 8 am, at lunch and at 8 pm
- Every call is recorded, transcribed and scored, so you finally see why people call
What to get right before the first call goes live.
These are the rules that most often catch businesses answering the phone with AI. This is practical guidance, not legal advice. Ask your counsel about your state and your industry.
Tell callers it is an AI9
Utah requires disclosure when a consumer clearly asks, and upfront for high-risk interactions that both collect sensitive data and give advice people rely on (law sunsets July 2027). Saying it in the first sentence of every call avoids the question.
Announce recording at the start10
Thirteen states require all parties to consent to recording, including California, Florida, Illinois and Washington. Federal law needs one party. Verbal notice at the start is the common method. Apply the strictest state.
Control how the vendor uses call audio11
California privacy suits claim an AI vendor that could use call data for its own purposes is an eavesdropper. Courts are split. Require written terms that limit use to your service and no model training.
Treat call-backs as outbound calls12
A call-back is a call you place. An inquiry gives a 3-month business relationship under the FTC rule, but prerecorded calls need written consent and the FCC treats AI-generated voices as artificial voices under the TCPA. Honor any do-not-call request. Get permission on the inbound call.
Keep health details off messages13
HHS says providers may leave short messages on answering machines but should limit what is revealed, such as name, number and appointment time. Do not put diagnoses or results in a text or voicemail.
Four questions to ask any voice AI vendor, including us
- Do you use our call audio or transcripts to train any model, yours or a third party's?
- Can we set the disclosure and recording wording, and can a caller reach a person at any point?
- What happens to a transfer if no one answers, and can the call fall back to a message?
- Where are recordings stored, how long, and can we delete one call on request?
This is general information, not legal advice. Rules differ by state and country and change often. Confirm with your own counsel before you call.
What an AI receptionist connects to.
VoiceGenie works with the carrier and numbers you already have. It reaches your calendar and CRM directly for common tools, and through the open API, webhooks or Zapier for the rest.
Phone systems and carriers
- Any SIP trunk or PBX
- TWTwilio
- PLPlivo
- TETelnyx
- VOVonage
- RIRingCentral
- MTMicrosoft Teams Phone
Calendars and scheduling
- GCGoogle Calendar
- MOMicrosoft Outlook
- CACalendly
- ACAcuity
- SASquare Appointments
CRM and practice or job systems
- HUHubSpot
- SASalesforce
- ZOZoho
- CLClio
- SEServiceTitan
- JOJobber
Where messages land
- SMS
- SLSlack
- MTMicrosoft Teams
- ZEZendesk
One platform for the AI, the dialer and your people, set up once
You do not need one vendor for the AI that answers, another for the dialer, another for the human agent desk and a fourth for the tools that help your staff. VoiceGenie runs all four. You connect your calendar, your CRM and your knowledge once. Each correction and each QA finding then improves the AI and your people together.
Four tools, four set-ups
own training · own reports
own training · own reports
own training · own reports
own training · own reports
One platform, set up once
- Your phone numbers and carrier
- CRM and calendar connections
- Knowledge and call rules
- Recordings, QA and analytics
Five ways AI receptionist projects go wrong.
Questions people ask before they start
What does an AI receptionist do?
An AI receptionist answers your main line, says it is an AI, and handles the routine calls: hours, directions and prices from your knowledge base, routing to a named person, taking messages, and booking or rescheduling. When a caller needs judgment, it transfers to a person with a summary, the transcript and what it already did.
How much does an AI receptionist cost?
VoiceGenie's AI is $0.14 per minute, billed per second, and the phone line is not included. A thousand minutes is $140. Human seats, if you add them, are $29, $59 or $99 per person per month on annual billing. For comparison, one national human answering service lists plans from $250 a month for 50 minutes.
AI receptionist vs answering service vs in-house receptionist: which is better?
They suit different jobs. An in-house receptionist handles walk-ins and judgment calls. A human answering service adds coverage but works from a script. An AI covers every call, at any hour, at a low per-minute cost. Most offices do best with the AI answering first and their people taking the calls that need them.
Will callers hate talking to an AI?
Some will, and design matters. Gartner found 64% of customers would prefer companies not use AI for service, but 87% say a way to reach a human is essential. In the CallRail survey, 66% would use AI to check hours and 50% to book. Keep the AI to clear jobs and make a person easy to reach.
How long does it take to set up?
Planning range: a few days to a week for an after-hours forward with a knowledge base and message delivery, and longer if you connect a CRM or complex scheduling rules. Your carrier or phone system sets the forward, and you keep your number. Plan two to three weeks of daily call review before widening.
Do I have to tell callers it is an AI?
There is no single US federal rule for inbound calls, but state laws are growing. Utah requires disclosure on a clear request and at the start of high-risk calls. The safe, simple practice is to say in the first sentence that this is an AI assistant and the call is recorded, and to offer a person on request.
Can it work with my current phone system and number?
Yes. VoiceGenie works with your own carrier and numbers, including Twilio, Plivo, Telnyx, Vonage and any SIP trunk or PBX. You forward calls to it after hours, after a set number of rings, or from one menu option. Your main number does not change, and you can switch the forward off at any time.
What happens with emergencies and angry callers?
They go to a person. You define the urgent situations and words, such as a gas smell, water damage or a medical symptom, and the AI tells the caller to call 911 for danger to life, then transfers to an on-call person with the details. A caller who asks for a human, or sounds upset, is handed over right away.
Related guides
From the VoiceGenie blog
Hear it handle your calls before you change anything.
Talk to our AI agent in your browser, or start a trial and point one line at it. Your team keeps every call that needs them.
No credit card · No minimum commitment · Keep your carrierSources
- CallRail, Why businesses can't afford to miss calls (2025).
- Invoca, How much missed sales calls cost home services businesses (2024).
- CallRail via Business Wire, CallRail Releases Report Benchmarking Marketing Efforts for Small Businesses (2025).
- CMSWire, reporting the Five9 2026 CX Report, Customers: the AI handoff is broken, contact center leaders won't hear it (2026).
- ContactBabel, US Contact Center Decision-Makers' Guide 2025 (2025).
- ServiceTitan, Quarterly industry trends, State of the Trades (2026).
- US Bureau of Labor Statistics, Receptionists: Occupational Outlook Handbook (2025).
- US Bureau of Labor Statistics, Employer Costs for Employee Compensation, Q2 2026 (2026).
- Davis Polk, Utah scales back reach of generative AI consumer protection law (2025).
- King and Spalding (KTS Law), Wiretap laws in the United States (2024).
- Epstein Becker Green, Eavesdropping by algorithm: how courts are applying decades-old privacy laws to AI (2026).
- US Federal Trade Commission, Q&A for telemarketers and sellers about DNC provisions in the TSR (2026).
- US Department of Health and Human Services, May health care providers leave messages? (2023).
- Gartner, Gartner survey finds 64% of customers would prefer that companies didn't use AI for customer service (2024).
- Gartner, Gartner survey finds 87% of customers say companies using GenAI must provide access to a human agent (2026).
- Ruby Receptionists, Virtual receptionist pricing (2026).