For most small walk-in businesses a phone AI receptionist is enough, and the only counter kiosk worth considering is a plain tablet check-in. Phone AI catches calls you miss while serving people; a counter screen helps only if walk-ins queue, arrive when nobody is at the desk, or need forms signed. Tally a week of both first.
A video-greeting kiosk is a different animal. One kiosk vendor's own pricing page starts its greeting plans at $299 a month before hardware, against a free tier to about $150 a month for phone AI and free to about $100 a month for a tablet check-in tool. The kiosk has to win back far more than the other two before it pays, so get written prices for all three before you decide.
Three jobs at the front of a walk-in business
Front-of-house work splits into three jobs, and each has a different tool. Catching calls when nobody can pick up is the phone AI's job. Greeting and checking in people who walk through the door is the counter screen's, which can be anything from a plain tablet to a video-greeting kiosk. Answering the same half-dozen questions, such as prices, hours, trial slots and wait times, can be done by the phone AI, a website chat or a printed sign.
| Tool | Does well | Can't do | Cost evidence |
|---|---|---|---|
| Phone AI receptionist | Answers when the desk can't, books or takes details, handles repeat questions at any hour | Greet someone standing in your shop; deal with a distressed or complex caller without a hand-off | A free tier to about $150 a month at the list prices below |
| Tablet check-in and waitlist | Takes names, adds people to a queue, texts when ready, collects forms | Answer open questions or judge urgency | Free to about $100 a month on the vendor pages below |
| Video-greeting kiosk | A screen with a face or voice that greets visitors and calls the right person | Replace a person who must judge a situation; run without good power and network | One vendor's greeting plans start at $299 a month, plus hardware |
| Website chat or a sign | Answers common questions before anyone calls or visits | Take a phone call or greet anyone | Varies; often free to start |
Tally one week before you spend anything
Keep a clipboard at the desk and a tick sheet by the phone. One week gives a hint, and four weeks smooths out a quiet or busy stretch, so run it for four if you can. You need nine counts, and nothing on the sheet should identify a customer. Here is a filled-in tally for a music school with 85 students (illustrative):
| Count | Four-week total |
|---|---|
| Calls received | 160 |
| Answered live | 124 |
| Missed during desk hours | 36 (22.5%) |
| Of those, calls from new families | 5 |
| Calls outside desk hours (voicemail) | 20 |
| Calls that were one of six standard questions | 88 |
| Walk-ins | 96 |
| Walk-ins when nobody was at the desk | 16 |
| Of those, new families who left without speaking to anyone | 2 |
Read it in two halves. The phone half asks how many calls from new customers go unanswered and what one new customer is worth. The counter half asks how often someone arrives and finds nobody, and whether they wait, wander off or are existing customers who don't mind. A tool is worth a trial when its monthly cost divided by the value of one new customer comes to less than the number of new customers it could plausibly recover each month. If a tool costs $100 a month and a new customer is worth $800, it pays for itself by recovering one customer every eight months.
A music school worked through: phone AI, tablet or kiosk
The school's front desk is staffed for two hours in the morning and three in the afternoon, and the director teaches the rest of the day. Assume an enrolled student pays $90 a month and stays nine months, so is worth $810. Assume 80% of missed new-family calls are captured as bookings or enquiries by a phone AI, and that one in eight captured contacts ends up enrolled. All of these are assumptions you'd replace with your own.
The tally shows 5 missed new-family calls and 2 unattended new-family walk-ins in four weeks. Phone AI recovers 4 of the 5 calls, which at one in eight is half an enrolment a month, worth $405. Its cost depends on the plan. The AI would take the 36 calls missed in desk hours plus the 20 out-of-hours calls, 56 in all. On a plan with 25 free calls and $3 for each after that, the bill is $93 a month. On a $150 flat plan it is $150.
| Option | Monthly cost | Extra enrolments a year to break even | Fit with the tally |
|---|---|---|---|
| Phone AI, free tier then per call | $93 | 2 (1.4 exactly) | Expected recovery is about 6 a year |
| Phone AI, flat plan | $150 | 3 (2.2 exactly) | Still comfortably ahead |
| Tablet check-in (assumed mid-range) | $30 | 1 (0.4 exactly) | Cheap; mostly saves desk time and collects trial forms |
| Video-greeting kiosk, cheapest listed greeting plan | $299, before hardware | 5 (4.4 exactly) | Expected recovery is about 3 a year, so it can't pay even before the screen |
The verdict for this school: start phone AI on the free tier, add a tablet only if parents queue or trial-lesson forms pile up, and skip the video kiosk. The unattended walk-ins are mostly existing parents who wait a few minutes, and only 2 a month were new families. A screen can't create those students; at best it stops two of them leaving. Notice too what the 31 desk-hour calls missed from current families show: the AI's job there is to take a message and flag urgent ones, and a person still has to deal with them.
The tablet's case rests on desk time, not enrolments. Suppose the director spends 15 minutes a day on sign-ins and forms, and that time is worth $30 an hour. That is $7.50 a day, or about $158 over 21 teaching days. A $30 tool that halves the job saves about $79, so it clears its cost by around $49 a month on time alone, but only if parents actually use it. If they walk past the screen to find a person, the saving vanishes and you are paying $30 for a prop.
For other tools in this area, the tutorials on what an AI receptionist costs a small business and on measuring an AI receptionist's return in 90 days go deeper on the phone side.
Where phone AI earns its keep, and where it doesn't
Phone AI pays when your hands are busy, when calls arrive outside opening hours, when most calls repeat the same questions, and when each captured enquiry is worth real money. It struggles with distressed or angry callers, complicated quotes, noisy backgrounds, and callers who want the owner and hang up on a machine. These examples show list prices on their vendors' pages as of October 2026:
| Product | Built for | List price |
|---|---|---|
| Smith.ai AI Receptionist | Any small business | Free for 25 calls, then $3 a call; Pro from $150 a month; two languages only |
| Fresha AI Concierge | Salons that use Fresha | $29.95 a month for messaging, or $99.95 for calls (200 minutes, then $0.60 a minute) |
| Booksy AI Receptionist | Salons and barbers that use Booksy | Beta; phone only; free for the first 500 providers |
| Jobber Receptionist | Trades that use Jobber | $29 a month add-on, included on the Plus plan; check how many conversations it covers |
| Yelp Host | Restaurants | From $249 a month after a trial of up to 30 days; transfers calls |
Per-call pricing makes volume the thing to watch. On a plan with 25 free calls and $3 a call after that, 40 calls a month cost $45, 56 cost $93, and 75 would cost $150, the same as the $150 starting price of the flat plan. So above roughly 75 calls a month a flat plan starts to make sense, subject to how many calls it includes, which you should read off the plan page. Below that, the free tier is the cheaper way to learn whether the tool works.
Before you commit, make ten test calls: five of your most common questions, one emergency, one angry caller, one who must spell a surname, one from a noisy street, and one who simply asks for a person. One vendor lists 50 simulated scenarios a month for rehearsing. Set the rules in writing as well: the AI says it is an AI at the start, it never confirms a booking the calendar can't hold, and it hands off to a human on named triggers. The tutorial on call scripts and escalation rules for an AI receptionist has wording for all three.
A sample call, and the sheet that feeds it
This is the kind of exchange a well-set-up phone AI has with a new family (illustrative, with prices left as placeholders):
AI: Hello, you've reached [school name]. I'm an AI assistant. I can answer common questions, book a trial lesson or take a message for the team. How can I help?
Caller: Do you teach drums to an eight-year-old?
AI: Yes, we teach drums from age six. A trial lesson is 30 minutes. Would you like me to find a time?
Caller: What does it cost?
AI: A trial lesson is [price] and a regular weekly lesson is [price] a term. I can text you our price list.
Caller: Yes please, and can he start on Thursday?
AI: I can request a Thursday trial. Can I take your name and number, and your son's name? A teacher will confirm within one working day.
Check three things in a sample like this. The AI said it was an AI before anything else. It requested a slot rather than promising one, which is honest when it can't see a teacher's real availability. And every figure came from the knowledge sheet below, which is why the sheet needs a review date. A badly set-up version of this call could easily say "from age five" because last year's sheet said so, or offer a Thursday slot the drum teacher had blocked. Both are sheet and calendar problems, not voice problems, and test calls are where they should surface rather than live ones.
School knowledge sheet (reviewed [date])
Lessons: piano, guitar, drums, violin, voice. Ages 6 and over.
Trial lesson: 30 minutes, $[price], booked through the desk or online.
Term fees: $[price] per term, paid in advance. Instrument hire: $[price] a month.
Desk hours: [times]. Teachers cannot take calls during lessons.
Hand off to a person: any call about a child's safety or collection, a complaint, a refund,
or a caller who asks for a person.
Never: confirm a place, quote a discount, or discuss another student.A sheet this short is easier to keep right than a long one. Put the review date at the top, and add "update the sheet" to the checklist you run whenever fees or timetables change.
Where a counter check-in screen earns its keep
A counter screen pays in four situations. A queue, as at a barber or repair shop, where taking names and texting "you're next" frees the person at the desk. An unstaffed desk, as at the music school, where the director is teaching. Forms, such as waivers, consent sheets or trial-lesson details, that you would otherwise collect on paper. And visitors, such as contractors or tour guests, who need signing in. A plain tablet does all four. The video kiosk adds a greeting and a call to the right person, which matters mainly in a building where reception routes visitors to many different staff.
Tablet tools are cheap enough to trial. TablesReady has a free plan with self check-in and 150 text messages a month per location, and paid plans at $39 and $79 a month billed annually. waitlist.me puts a self check-in kiosk mode on its Premium plan at $34.99 a month ($27.99 billed annually), with 1,000 notifications a month and 2 cents for each one after that. Neither page makes clear whether every plan is billed per location, so ask before you add a second site.
Visitor sign-in systems sit a step up. Greetly lists $99 and $159 a month billed yearly and runs on off-the-shelf hardware. Video-greeting kiosks are a different bracket again: ALICE Receptionist lists its front-desk plan, with a motion-triggered greeting, live calls to staff and AI avatars, at $299 or $343 a month depending on billing, and says some features need extra hardware or one-time set-up fees. Ask for written quotes, and ask what happens when the network drops.
Most counter-screen failures aren't about the screen. They happen because nobody notices the notification. Send alerts to the phone or watch of whoever covers the floor, and test the gap with a stopwatch: from tapping "check in" to a staff member looking up. Plan for access too. Check the height, glare and font size, offer another way in for people who can't or won't use a screen, and keep a bell. The tutorial on AI customer service kiosks for a shop counter covers the question-answering kind of kiosk, and AI receptionists for barbers covers walk-in queues in depth.
Two more businesses: a kitchen showroom and a nursery
A kitchen fitter with a showroom
Walk-ins are few and mostly by appointment, so the screen has little to do. The phone is the business: the fitter is on site fitting units, and a missed enquiry call goes to a competitor. Say there are 18 enquiry calls a month and 6 go unanswered. If one enquiry in ten becomes a $9,000 job at a 25% margin, each enquiry is worth $225 in expectation, and the 6 missed calls represent $1,350 a month. A phone AI that recovers a third of them, 2 enquiries, is worth $450 against a $150 plan. A showroom visit is a conversation with a person, so a doorbell that alerts a salesperson does the screen's job.
A childcare nursery
A nursery's phone is mostly existing parents with urgent messages: a sick child, a late pickup. Use phone AI for new enquiries and tour bookings only, and route existing parents to staff during opening hours. Anything mentioning illness, injury, collection or safeguarding is an instant hand-off. A sign-in tablet for visitors and contractors is reasonable, but it must never decide who may collect a child: those checks stay with staff. The tutorial on AI chatbots for nursery enquiries and visit bookings covers the enquiry side.
Which tool for which business
| Business | Phone AI | Counter screen |
|---|---|---|
| Barber or other walk-in service | Yes, for "how long is the wait?" calls | A tablet waitlist, yes; it is the core of the job |
| Repair shop | Yes, since calls arrive while hands are busy | A tablet drop-off form; not a video kiosk |
| Small clinic | Only with strict hand-off rules | Check-in usually sits in the practice software |
| Music school | Yes, free tier first | Optional tablet; skip the video kiosk |
| Kitchen showroom | Yes: enquiries are the business | A doorbell and staff alert is enough |
| Childcare nursery | Enquiries and tours only | Visitor sign-in tablet; never collection checks |
| Cafe | Rarely worth it | No |
| Retail shop | For stock and hours calls | No |
Four rules of thumb from your tally, mine rather than a standard:
- Missed calls under about one in ten and no lost new-customer enquiries: do nothing yet, and repeat the tally in a quarter.
- Missed calls above one in ten, with new-customer enquiries among them: trial phone AI on a free or per-call plan.
- Walk-ins finding nobody at the desk at least weekly, or a queue that makes people leave: trial a tablet check-in.
- A video-greeting kiosk only if visitors must be routed to many different people and no one can staff the door.
Failure modes at the counter and on the line
- A tablet nobody answers. The notification goes to a computer in the back office while a customer stands in front of the screen. Fix: alerts to a phone or watch, and a stopwatch test.
- A trial booked in a blocked slot. The AI wasn't connected to the real calendar. Fix: connect it, or let it take requests rather than confirm bookings.
- Last term's prices. The AI quotes fees from an old document. Fix: one source document with a review date, updated whenever fees change.
- The emergency caller. A caller who mentions smoke, a collapse or a child in danger should hear "call the emergency services" first and be transferred second. Test it before you go live.
- Misheard names and numbers. Have the AI read details back and send a text confirmation. Check any booking made from a poor line.
- A screen at the wrong height or in glare. Test at different times of day and with someone who uses a wheelchair, or has low vision.
A 30-day test plan with a stop rule
| Week | Phone AI | Counter screen |
|---|---|---|
| 1 | Write the knowledge sheet, make the ten test calls, fix the gaps | Set up the tablet in a quiet corner; time the gap from check-in to a person looking up |
| 2 | Divert after-hours calls only; read every transcript | Run it beside the clipboard for a week |
| 3 | Add overflow during desk hours; listen to 20 calls | Drop the clipboard; compare check-ins with a hand count of walk-ins |
| 4 | Compare with the four-week baseline; work out the cost per captured enquiry | Same comparison; decide on the date you wrote down at the start |
For phone AI, count calls handled, the share it completed without a human, hand-offs, bookings made, and complaints. Listen to 20 calls if you record them and callers have been told, and repeat your ten test calls weekly in the first month. Compare the month's tally with your four-week baseline: missed calls from new customers should fall towards zero. If after 30 days it has handled fewer than 20 calls, volume or set-up is the problem, and you should decide whether to stop, not drift on at $150 a month.
For a counter screen, compare check-ins with a hand count of walk-ins, measure the time from check-in to a person looking up, and note any staff minutes saved. If fewer than half of walk-ins use it by week three, ask why before blaming customers. In both cases set a decision date at the start, write down the break-even number from your own tally, and stop on the date if the numbers don't support carrying on.
Phone AI and counter screen questions
Can one system do both the phone and the counter?
Some vendors sell a voice assistant for calls and a separate screen or kiosk, occasionally under one account. They are different jobs with different failure modes, so price them separately and ask whether each works without the other. Don't pay for a bundle to get one half you won't use.
Do I have to tell callers they are speaking to an AI?
Tell them at the start of every call; it is fair and it cuts complaints. If you have customers in the EU, the AI Act's transparency duty for systems that talk to people has applied since 2 August 2026, and rules on recording calls differ by place, so ask an adviser. Pick a tool that lets you switch recording off.
What if my customers are older and dislike talking to machines?
Offer an easy way out: a keypress or a spoken word that reaches a person, or a callback promise within the hour during opening times. Watch the hang-up rate in the first month. If many callers abandon, the AI is losing the customers you meant it to catch, so narrow its job to after-hours calls.
Is a free tier enough to test phone AI?
For a trial, often yes: one vendor's free plan covers 25 calls, then $3 each. That is enough for ten test calls and a week of real overflow. Check the free tier's limits on languages, hand-off and calendar links before relying on it, and set a fixed trial period with a decision date.
Further reads
- AI for a Music School: Lesson Notes and Parent Emails Worked Through — The same music school's back-office side: lesson notes and parent emails.
- How a Small Clinic Can Use AI at the Front Desk Safely — Front-desk AI where hand-offs and privacy matter more.
- AI Receptionist for Retail: Shop Cost, Stock Calls and Limits — What a phone AI costs a shop, and which calls it should take.
- Does a Coffee Shop Need an AI Receptionist? When It Pays — A case where neither tool usually pays.
- How to Set Up an AI Receptionist Without Losing Callers — Set-up steps that stop you losing callers.
- AI Receptionist or Answering Service for a Booking-Based Business? — Compare phone AI with a human answering service.
- What Is an AI Receptionist and How Does It Handle Bookings? — How an AI receptionist handles a booking, step by step.
- What to Do When an AI Receptionist Gets a Booking Wrong — What to do when the AI gets a booking wrong.
- Kiosks AI for Cafés and Takeaways: Self-Service Ordering Costs — Hardware, software fees, menu set-up and a break-even sum for a self-service kiosk, plus the cases where a person at the counter still wins.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Smith.ai, Fresha, Booksy, Jobber and Yelp pricing and feature pages; TablesReady, waitlist.me, Greetly and ALICE Receptionist pricing pages; all read in October 2026. Prices change and hardware is extra, so get written quotes.