To set up an AI receptionist without losing callers, let your own phone ring first and forward only unanswered, busy and after-hours calls to the AI after about 15-20 seconds. Give it a greeting under ten seconds that says it's an AI, a way to reach a person, and an automatic text-back. Then test 30 calls before launch.
Most lost callers never reach the conversation part. They disappear in the plumbing: a forwarding rule that sends calls to an old voicemail box, a welcome so long that people hang up before the AI asks anything, or a transfer that rings out with nobody there. Each failure is silent, because a caller who gives up doesn't complain; they ring the next salon on the list. So build the routing, fallbacks and tests first and polish the wording second. If you're still deciding whether you want one at all, read how an AI receptionist handles bookings before you start.
Three places callers slip away, and the trace each one leaves
Every dropped call leaves a small mark somewhere, even when nobody complains. Knowing where to look is half the setup.
| Leak point | What the caller experiences | How it shows up for you | What prevents it |
|---|---|---|---|
| The forward | Endless ringing, or an old voicemail box that's full | Fewer AI calls than your missed-call count; voicemails nobody listens to | Conditional forwarding for no answer, busy and unreachable, each one tested |
| The first ten seconds | A long welcome, a pause before the AI speaks, a voice that talks over them | Lots of calls under ten seconds; the same number ringing again within the hour | A greeting under ten seconds that gets to a question quickly |
| The handover | A transfer that rings out, or "someone will call you" and nobody does | Unanswered transfers; callbacks still open the next morning | A timed fallback on every transfer and a callback promise with a time on it |
Stage 1: pick which calls the AI answers (about 20 minutes)
The lowest-risk setup is the one where the AI only meets callers who would otherwise have got nothing. There are four common routing modes, and you can combine them.
| Routing mode | How it works | Risk of losing callers | Suits |
|---|---|---|---|
| After hours only | Calls outside opening hours go straight to the AI | Low: it never competes with your staff | Businesses that answer well by day but miss evening enquiries |
| No-answer overflow | Your phone rings for 15-20 seconds, then the AI picks up | Low, if the timer is right | Salons, studios and barbers whose hands are busy mid-appointment |
| Busy overflow | A second caller reaches the AI while you're already on the line | Low | Single-line businesses with bursts of calls |
| Every call | The AI answers first, always | Highest: regulars who want a person meet the AI too | Businesses that have run overflow cleanly for a month or more |
For a first setup, combine after hours with no-answer and busy overflow. A regular who rings while you're free still gets you, and the AI's mistakes can only land on calls that were already being lost. Move to answering every call later, if at all, once the logs show it copes.
Stage 2: wire the forwarding and prove every path (about an hour)
Forward, don't port, for the first month. Keeping your number with your current phone provider and forwarding calls to the AI's number means you can switch the AI off in seconds from your own phone. Moving (porting) the number to the AI provider is a bigger job that is slow to reverse, so leave it until the logs have earned it.
Mobile networks usually offer three conditional forwards: no answer, busy and unreachable (phone off or no signal). Set all three to the AI's number. On many networks the no-answer delay can be set in five-second steps up to 30 seconds, and your carrier's help pages list the exact dial code or menu. Set it to 15 or 20 seconds: long enough for someone free to answer, short enough that the caller hasn't given up.
Watch for the voicemail race. If your phone's own voicemail answers after 20 seconds and your forward waits 25, voicemail wins every time and the AI never hears a thing. Either shorten the forward or switch voicemail off on that line.
The vendors describe the same mechanics in their own terms. Jobber's help page for its Receptionist notes that most carriers can forward after a set number of rings or outside business hours, and it offers an optional 10-second answering delay, which it recommends when you forward through Google Voice. Quo (formerly OpenPhone) lets every plan forward all calls, but forwarding only unanswered calls based on business hours sits on its Business and Scale plans. Check your plan before you design around a feature it doesn't have.
Then prove it with four calls from a second phone:
- Let it ring out: the AI should pick up at your chosen delay.
- Decline the call: the busy forward should send it to the AI.
- Switch your phone off and call: the unreachable forward should catch it.
- Call ten minutes after closing: the after-hours route should answer with the closed greeting.
Here's the mistake this catches. In one illustrative nail salon, the owner set no-answer forwarding on the salon mobile but skipped the unreachable rule. On the Sunday the phone was switched off, eleven calls went to the network's default voicemail, which was already full. The AI dashboard showed a quiet Sunday. The gap only surfaced on Tuesday, when two regulars mentioned they'd tried to book.
Stage 3: write a greeting nobody hangs up on (about 30 minutes)
Callers decide within a few seconds whether to stay. A greeting that sounds like a hold message loses them before the AI asks a single question. Compare these two, read aloud with a stopwatch.
Before (about 22 seconds):
"Hello and thank you for calling [salon name], where we're passionate about giving you perfect nails every time. Please note that this call may be recorded for quality and training purposes. I am an AI assistant and I can help you with bookings, prices, opening hours, gift vouchers and much more. How may I help you today?"
After (about 8 seconds):
"Hi, you've reached [salon name]. I'm the salon's AI assistant while the team's with clients. I can book you in, answer questions, or take a message for a person. What can I do for you?"
The shorter version does five jobs in order: it confirms who they've reached, says it's an AI, explains why a person isn't answering, shows there's a human route, and hands the turn to the caller. The list of services disappears because callers will tell you what they want.
Set separate open and closed greetings; most tools support this, and Jobber's version is a business-hours toggle. The closed greeting should say when you reopen and offer to book or take a message. Some tools can greet recognised numbers by name, which feels warm until a shared family phone greets a husband by his wife's name, so test it with a few real client records before you rely on it.
Two notices belong in the greeting only if they apply. If you serve customers in the EU, the EU AI Act's transparency duty for chatbots has applied since 2 August 2026, so callers must be told they're dealing with an AI; the "I'm the salon's AI assistant" clause covers it. If the tool records calls, say so briefly. Rules on recording consent differ from place to place, so check yours with an adviser rather than copying another business's wording.
Stage 4: a one-page fact sheet beats pointing it at your website
Allow one to two hours for this stage. Most tools will happily read your website, but don't rely on that alone. Websites carry last year's prices, a promotion that ended in March, and marketing lines the AI will repeat as promises. Give it one page it treats as the truth, dated, and update it the day anything changes. Here's one filled in for an illustrative three-technician nail salon.
| Section | Nail salon example |
|---|---|
| Services and durations | Gel manicure 45 min; gel removal and new set 75 min; acrylic full set 90 min; infills 60 min |
| Prices it may quote | Gel manicure $38; removal and new gel set $52; acrylic full set from $55; infills from $40. Always say "from" for acrylics and never quote nail art |
| Booking rules | No new bookings within 30 minutes of closing; new acrylic clients get a 90-minute slot; Saturday bookings need a $10 deposit, sent by text link |
| Cancellations | 24 hours' notice, or the deposit is kept |
| Walk-ins | Welcome Tuesday to Thursday if the calendar shows a free chair; the AI checks before promising |
| Never | Comment on reactions, infections or nail damage; promise a named technician the calendar doesn't show free; offer discounts |
| Hand to a person | Complaints, any reaction or injury, groups of four or more, press, suppliers and job applicants |
The "never" row matters as much as the prices. When a booking does go wrong, the fix is usually in this sheet, and what to do when an AI receptionist gets a booking wrong covers the clean-up with the client.
Stage 5: build the escape routes before a caller needs one
Budget about 45 minutes. An AI receptionist loses callers fastest when it can't let go of them, so build four routes out.
- Transfer on request. Anyone who asks for a person, or says a trigger word such as "complaint", "reaction" or "manager", should be offered a transfer within two turns. Jobber's Receptionist can listen for keywords you choose and either text you or transfer the call. Quo's Sona offers an Escalation job template built around a Transfer call action.
- A fallback when the transfer isn't answered. If Sona transfers to a Quo user who doesn't pick up, the call goes to voicemail, so record a voicemail greeting with a real next step: "Leave your name and we'll call you back before 6pm." A transfer that rings out into silence is the worst outcome of all.
- A text when callers hang up early. Jobber's Text Back option texts callers who disconnect before the conversation starts. If your tool has nothing similar, add a missed-call text-back so early hang-ups still get a reply.
- A plan for running out of allowance. Quo's Sona pricing page says overages are switched off by default and, once credits are used up, calls go to the fallback path. Jobber's add-on charges $0.79 per conversation after the first 30. Goodcall charges $0.50 per unique caller over the plan. Decide now whether you'd rather pay the overage or send those calls to a person, because the default may not be the one you want.
Every message the AI takes should promise a time you can actually meet. "Someone will call you back before 6pm today" earns trust; "as soon as possible" invites a second call to a competitor. The exact wording for each escalation, filled in for a tattoo studio, is in call scripts and escalation rules for an AI receptionist.
Stage 6: thirty test calls before any customer hears it
This takes about two hours. Run ten scenarios with three different testers: you, a colleague, and a friend with a different voice, pace or accent. Use their own phones, not yours, so caller ID behaves like a stranger's. Mark each call pass or fail against the right-hand column.
| Scenario | What the tester says | Passes if |
|---|---|---|
| Standard booking | "Can I get a gel manicure Saturday around eleven?" | Correct service length, a real free slot, confirmation text, entry in the calendar |
| Price with a condition | "How much to take off my gels and do a new set?" | Quotes $52 from the sheet and doesn't invent extras |
| Asks for a person | "Can I just talk to someone?" | Transfer offered within two turns, or a message with a callback time |
| Noisy line | Calls from a busy street or with a hairdryer running | Asks them to repeat rather than guessing a service |
| Fast or accented speech | Rapid booking request with a name that's hard to spell | Spells the name back and confirms the number |
| Hangs up mid-booking | Ends the call before confirming | No half-made booking left behind; a text follows within a minute |
| After closing | Calls at 9:30pm | Closed greeting, then offers a booking or a message |
| Complaint | "My nails lifted after two days" | Apologises, offers no refund, escalates to a person |
| Reaction or injury | "My skin's gone red and itchy since Thursday" | Suggests seeing a medical professional, escalates at once, gives no product advice |
| Silence or robocall | Says nothing, or plays a recorded sales message | Ends politely within about 20 seconds |
Anything that fails twice goes back to stage 4 or 5 before launch. Keep the recordings: they're the baseline you'll compare the first real week against.
A three-technician nail salon's first month, in numbers
Here is how the arithmetic can look, using made-up but realistic figures. The salon takes about 180 calls a month. Before the AI, 54 of those rang out during opening hours (30%, mostly between 11am and 4pm when all three technicians had clients) and another 16 came in after closing. That's 70 calls a month reaching nobody.
The owner set after-hours plus no-answer overflow at 20 seconds, and busy overflow. In month one the AI answered those 70 calls. Six were silent or spam, leaving 64 real conversations:
- 30 bookings made directly in the calendar
- 13 questions answered (prices, parking, whether they do children's nails)
- 8 messages taken, each with a "before 6pm" callback promise
- 8 transfers: 5 answered at the front desk, 3 rang out between 1pm and 2pm and dropped to voicemail with a text-back
- 5 hang-ups within ten seconds, just under 8% of conversations
At an illustrative average ticket of $45, the 30 bookings are worth $1,350. Assume half those callers would have rung back anyway and the recovered revenue is still about $675 for the month.
Against that, here's what 64 conversations would cost on three tools at the list prices in the table further down. Goodcall's Starter plan is $79 a month on monthly billing, since roughly 55 unique callers sits inside its 100-caller allowance. Smith.ai's free AI plan includes 25 calls, so 39 more at $3.00 comes to $117. A salon already on Quo for its phones would need Sona's $49 tier of 10,000 credits (about 100 calls) on top of the Quo plan. The cheapest option at this volume isn't the one with "free" in its name. For a fuller cost picture, see what an AI receptionist costs a small business.
The month's real lesson came from the three failed transfers. All of them happened during the front desk's lunch hour, so the owner pointed transfers for that hour to her own mobile. The next month had none.
What the first fortnight of call logs should tell you
Spend 15 minutes a day for the first two weeks on the log and five recordings. You're looking for these patterns:
- Early hang-ups above about 15% of answered calls. The greeting is too long, or there's a delay before the AI speaks.
- The same number calling twice within an hour. The first call failed somewhere; listen to it.
- More than two "could you repeat that?" turns in one call. Background noise or accents are beating it; add a line offering to text the caller a booking link instead.
- Bookings with the wrong length. Your fact sheet's service list is ambiguous.
- Unanswered transfers clustered in one hour. Move the transfer target for that hour.
- Callbacks later than promised. Change the promise, not the AI.
Check how long your tool keeps call history. Goodcall's Starter plan keeps seven days and Growth keeps 30, so on a short retention window, review or export weekly or the evidence disappears. After the fortnight, switch to a weekly review and track the return properly with a 90-day measure of the AI receptionist's return.
Four receptionist tools and how each one fails over
The right tool is partly the one whose limits you can live with. These are list prices as of September 2026; check each pricing page before you commit, because they change often.
| Tool | Price basics | At the limit | Route to a person |
|---|---|---|---|
| Jobber Receptionist | $29 a month add-on for 30 conversations, then $0.79 each; included with unlimited use on Jobber's Plus plan | Keeps answering and bills extra conversations; spam and "no conversation" calls aren't counted | Keyword triggers that text you or transfer the call |
| Quo Sona | 1,000 credits a month included (about 10 calls at 100 credits each); add-on tiers from $25 to $199 a month; calls under 15 seconds aren't counted | Overages off by default; calls go to your fallback path once credits run out | Escalation job with a Transfer call action; unanswered transfers to Quo users go to voicemail |
| Smith.ai AI Receptionist | Free plan with 25 calls, then $3.00 each; Pro from $150 a month for 75 calls | Extra calls billed per call; known spam calls filtered out | Human receptionists available for transfers on all plans |
| Goodcall | Starter $79 a month per agent on monthly billing ($66 annually) for 100 unique callers; $0.50 per extra caller | Keeps answering; per-caller overage | Check transfer options on your plan; Starter includes one logic flow and seven days of call history |
Jobber's receptionist makes most sense if you already run your jobs in Jobber, and Sona if your phones already sit on Quo. Smith.ai suits businesses that want a human safety net behind the AI. Goodcall's per-caller pricing is easy to predict for a salon whose regulars ring more than once a month.
Setting up an AI receptionist: follow-up questions
Should I move my business number to the AI receptionist provider?
Not at first. Forwarding keeps the number with your current phone provider, so you can switch the AI off in seconds from your own phone if something goes wrong. Porting the number to a new provider is a bigger job that is slow to reverse. Run on forwarding for at least a month, read the call logs, and only consider porting if a feature you genuinely need depends on it.
Do I have to tell callers they are talking to an AI?
It is the safest default everywhere, and in some places it is required. If you serve customers in the EU, the EU AI Act's transparency duty for chatbots has applied since 2 August 2026, so people must be told they are dealing with an AI. Callers who only realise halfway through tend to feel tricked. One short clause in the greeting, such as 'I'm the salon's AI assistant', covers it without slowing the call.
What do callers hear if the AI receptionist service goes down?
That depends on your forwarding rules and the fallback settings in the tool, so find out before it happens. Pause the AI and call your number from another phone. If the call rings out or reaches a full voicemail box, fix that first: record a current voicemail greeting with a callback time, and check it daily for the first month. Some tools let you set a fallback path in the call flow for exactly this case.
How long does a proper AI receptionist setup take?
Plan on roughly a working day spread over a week: 20 minutes to choose routing, an hour to set and test forwarding, half an hour on the greeting, one to two hours on the fact sheet, 45 minutes on escalation and fallbacks, and two hours for 30 test calls. Then allow 15 minutes a day for the first fortnight to read logs. The fact sheet and the testing take the time, not the technology.
Further reads
- AI Receptionist or Answering Service for a Booking-Based Business? — Weigh an AI line against a human answering service for bookings.
- Can AI Answer My Business Phone? What It Can and Can't Handle — What phone AI handles well and where it still struggles.
- How to Set Up an AI Phone Agent for After-Hours Calls — If you only want the AI answering once you have closed.
- Best AI Receptionists for Salons, Compared by Price — Salon-focused receptionist tools compared by price.
- AI Call Summaries: Log Every Phone Enquiry in Your CRM — Log every AI-handled call in your CRM automatically.
- How to Reduce No-Shows With AI Reminders and Automatic Rebooking — Protect the bookings the AI takes with reminders and rebooking.
- Customer-Facing or Back-Office: Where Should AI Go First? — A side-by-side comparison of customer-facing and back-office AI as a first move, with a scoring sheet, the middle route, and two worked decisions.
- Pet Grooming Software With AI: Features and Prices Compared — What the AI in MoeGo, Teddy, Groomify, DaySmart Pet and Gingr actually does, what each costs a year, and a 14-day trial script for groomers.
- AI Receptionist vs Front-Desk Hire: The Real Cost for a Salon — An AI receptionist costs $50-$250 a month; a desk hire costs thousands. But the desk does more than phones. Here's how to compare them fairly.
- How Dry Cleaners and Laundries Use AI for Orders and Collections — Where AI fits between the counter and collection: phone agents that book pickups, ready texts, an uncollected-garment ladder and calmer claim replies.
- Should a Hair Salon Use an AI Receptionist? — A decision test for salon owners: count your missed calls, check your software, and set the colour and patch-test rules before any AI answers the phone.
- What Does an AI Receptionist Cost a Dental Practice in 2026? — Published AI receptionist prices, every line on a dental practice's bill, three budgets by practice size, and the questions that expose the real quote.
- AI Receptionist for Tradespeople: What It Costs and When It Pays — Current prices for AI receptionists used by trades, how the billing units differ, and a payback sum based on the jobs you miss on the tools.
- Should a Small Auto Repair Shop Use AI? A Decision Checklist — A checklist for small repair shops deciding on AI: which problems justify it, what your shop system already does, costs, trust and a scoring rule.
- Setting Up an AI Answering Service That Never Misses an Emergency — Emergency definitions, a word-for-word screening script, a transfer chain with backups, and monthly test calls that prove your AI line catches urgent calls.
- How Independent Pharmacies Use AI to Shorten Phone Queues — Count the calls, then cut them: ready-to-collect texts, a phone menu that listens, a voice assistant with firm limits, and callbacks instead of hold music.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Jobber help article, Receptionist powered by Jobber AI; Quo pricing page, plus the Sona pricing and Sona job guide pages on Quo's support site; Smith.ai AI Receptionist pricing page; Goodcall pricing page; EU AI Act Article 50 transparency duty. Checked September 2026.