AI can take much of the repeatable message and phone work in a nail salon receptionist's job description: opening hours, price and service questions, booking, rescheduling, reminders and first-draft review replies. A person still has to greet walk-ins, take payments, handle complaints, judge health concerns and make exceptions. Expect it to trim desk hours, not replace the desk.
The split is yours to set, duty by duty, because it depends on your booking software, how many walk-ins you take and how much a mistake costs you. A duty moves to AI only when the answer exists in writing, a wrong answer is cheap to fix and the handover to a person is clear. A duty that fails any one of those three tests stays with a human.
Three tests before a duty moves to AI
Run every duty through the same three questions. Does the answer exist in writing? Opening hours, the price list, the cancellation policy and the parking instructions do. "Can you copy this photo in an hour?" does not, because only a technician can judge it. Is a wrong answer cheap to fix? A clash in the diary that a person spots at the start of the day is cheap; a price quoted to a client that you then feel bound to honour is not. Is the handover clear? Someone specific must take over, quickly, with everything the client has already said.
Here is a Saturday inbox sorted that way (illustrative). At 9:15 there were 14 new messages. Five asked about prices or hours, three asked to book a standard gel manicure in a free slot, one asked to move a booking and one asked whether the salon takes deposits. Those ten pass all three tests. Two were inspiration photos asking for a quote, one asked for a refund after a lifted nail, and one asked whether a treatment is safe in pregnancy. Those four fail: the quote needs a technician's eye, the refund involves money and the pregnancy question is about health.
That ten-from-fourteen split is deliberately generous. Calls run longer and messier than messages, and a Monday inbox looks different from a Saturday one, so the time log further down assumes a lower share. The way to find your own figure is a tally: for one week, have the receptionist mark every enquiry as "answer exists in writing" or "needed a person".
The duty-by-duty split for a nail salon front desk
The table lists fifteen duties a nail salon front desk typically covers. Read it as an owner's view, not an advert: the left column is the duty, the next two say who does what, and the last says when the AI must stop and a person must start.
| Duty | What AI can do | What a person keeps | Hand over when |
|---|---|---|---|
| Answer calls and messages about hours, address, parking, prices, services | Answer from the written menu and a knowledge base; send the booking link | Keep the knowledge base current when prices or hours change | The question is not in the knowledge base, or the client asks for a person |
| Book, move and cancel standard appointments | Book services whose length and price are set in the booking software | Custom or long services, groups, overrides | More than one technician is needed, three or more people, or an add-on is not on the menu |
| Reminders, confirmations, deposit links, waitlist offers | Send them on a schedule (often plain automation rather than AI) | Waive deposits; rule on late cancellations | A client disputes a deposit or no-show fee |
| Greet and check in clients at the door | Nothing beyond a QR or tablet sign-in | The greeting, the seat, the first impression | Always a person |
| Walk-ins | State the walk-in policy; give a wait only if fed live availability | Decide who sits where and run the queue | The client is present, in a group, or has access needs |
| Payments, tips, gift cards, packages | Explain how gift cards work; link to the purchase page | All card and cash handling, refunds, discounts | Always a person |
| Complaints, redo and refund requests | Apologise for the experience, collect facts, escalate | Decide the outcome and reply | Immediately |
| Allergic reaction, infection, injury, health questions | A safety message and an instant alert to the owner | Everything else, including the incident record | Immediately |
| Review replies and social messages | Draft replies in your tone; answer simple questions | Approve and post anything about a complaint | A review mentions injury, hygiene, a named staff member or a refund |
| Aftercare and retail | Draft aftercare messages after each visit | Recommend products to a client in the chair | A client asks about a reaction rather than a product |
| Supplies and stock | Turn the weekly usage count into a reorder list | Count, order, check deliveries | A supplier issue or an unusual price change |
| Hygiene support: towels, restocking, reset checks | Format and review the daily logs | Do the tasks and sign the sheet | A log shows a short soak or a missing entry |
| Client forms and consent records | Send forms and chase missing ones | Read the answers and decide whether a service goes ahead | A form mentions an allergy, a skin condition or a medical note |
| End-of-day sales summary | Summarise the booking or till export | Count cash, bank it, sign off | The totals do not match the cash |
| Staff diary and breaks | Draft rota options for the week | Decide, and talk to the team | Leave requests or a sickness gap |
The hygiene row ties to the daily hygiene logs tutorial, which shows what the person at the desk would be signing.
What the booking apps' AI does today, and where it stops
Many nail salons already pay for a booking app, and the AI worth testing first is usually the one inside it. As of October 2026, from each vendor's own pages:
| Tool | Channels | What it does | Handover | Cost and limits |
|---|---|---|---|---|
| Fresha AI Concierge | Calls, client messages and review replies | Books, reschedules and cancels in the Fresha calendar under your cancellation and rescheduling policies, telling the client first if a fee applies; answers common questions from a knowledge base you fill in | Passes calls to a transfer number you set, or takes a message for the team; you choose which request types always transfer | $29.95 a month per location for messaging only; $99.95 for calls and messaging, including 200 minutes per location, then $0.60 a minute |
| Booksy AI Receptionist (beta) | Phone calls only | Books new and returning callers against your menu and live availability; filters sales calls | Sends a summary and a callback request; full transcript in the activity log | By request while in beta; free for the first 500 providers for at least 12 months; not yet for appointments with No-Show Protection; English and Spanish only |
| Vagaro Vera | Vera Receptionist is set up as a chatbot; the pricing page does not mention calls | Described as helping to automate marketing and fill empty appointment slots; check which bookings it completes before relying on it | Check in your account | Listed under AI tools included with every plan; the Vera Fill My Books add-on says fees apply, and Vagaro's help centre calls Vera Receptionist free for a limited time |
Two points about those rows. Handover is a feature you configure, not one you get for free: Fresha says you choose when calls are handled by AI and when they are transferred to you, so the written split above is what you type in. And Fresha's tool can reply to reviews, either publishing on its own or waiting for you to approve each reply, so pick that setting deliberately. My rule is that nothing goes out on a one- or two-star review without the owner's eyes on it. If you want the detail on one tool, there are separate tutorials on Fresha AI Concierge and on what Vagaro's Vera does and where it stops.
Walk-ins: the door no AI can open
A walk-in is a person standing at your door, and no AI can greet them. The AI's job starts earlier and ends sooner. Before the client arrives, it states your policy and, only if it is fed live availability, a realistic wait. Once the client is in the room, a person takes over.
Four rules cover most salons. Whoever is free greets within about half a minute, and the owner decides who that is on each shift. A tablet or QR code sign-in puts the client on a waitlist with a name, a service and a phone number. The person at the desk, not a program, decides who sits with which technician. And the AI never promises a time it cannot see. If the salon would like help with the wait itself, the tutorial on walk-in wait estimates and technician matching covers it.
Hand over at once when the client is already in the building, when the group is three or more, when someone has mobility needs, when a child is involved, or when a caller sounds distressed or aggressive. A first draft of a reply to a Saturday walk-in enquiry (illustrative) read: "Come on down, we'll fit you in within 15 minutes." Nobody could know that, so the edited version says: "Walk-ins are welcome when a technician is free, but Saturday afternoons are usually fully booked. Booking online takes a minute: [link]. If you are already outside, come in and the team will tell you the wait."
Complaints: what the AI collects and what only you decide
Sort complaints into three kinds, because they need different speeds. Quality: a lifted nail, an uneven finish, a chip within days. Money: a refund request, a charge or deposit dispute. Health and safety: a reaction, a sign of infection, an injury. At first the AI treats all three the same way: apologise for the experience, collect the facts, say who will reply and by when, and pass it on. For health and safety it also stops chatting and alerts the owner immediately.
The AI never admits fault, offers a refund or a redo, names a cause ("that sounds like an allergy to the gel"), blames a technician or argues. All of those are decisions, and decisions belong to you. A message that follows those rules (illustrative):
I'm sorry this happened. I can't decide on a fix myself, so I'm passing it to [owner name] with the details you've given me. You will hear from a person by [time]. If your skin is sore or swollen, or you feel unwell, please see a doctor or pharmacist.The handover note is what saves the owner ten minutes of back-and-forth. It carries six fields, shown here filled in:
HANDOVER: complaint (quality)
Client: [first name], phone on file
Booking: Tue 3 Nov, 14:00, full set with chrome, technician [name]
What happened, in the client's words: "Two nails lifted at the edges after five days."
Photos received: yes, three
Asked for: a fix, not a refund
Urgency: same day. No health concern mentioned.
AI told the client: a person will reply by 17:00 today.Decide your reply window in advance and write it into the instructions. "By 17:00 today" only works if someone is actually checking at 16:30. If nobody can promise a time, the AI should say "within one working day" and mean it. For wording that suits your voice, the tutorial on call scripts and escalation rules for an AI receptionist (listed at the end) has a longer set of lines to adapt.
Inspiration photos and the quote problem
Nail clients send photos: "Can I get this?" The quote depends on length, shape, finish, hand-painted art, charms and how many nails carry them, and a wrong guess costs a technician an hour. The risk in the first month is a booking assistant that treats every photo as a standard manicure.
A tier table fixes most of it. Write three or four tiers with the duration each adds and which add-on lines from your price list apply, then give the table to your own assistant (not the booking app's) and ask it to match a photo to a tier. For example (illustrative): "Tier 1, one colour or one accent nail, adds no time. Tier 2, chrome or ombre across all nails, adds 30 minutes. Tier 3, 3D charms or hand-painted art on three or more nails, adds 60 minutes and needs technician approval." The assistant proposes a tier and a slot length; a technician confirms before the booking is made.
Here is how it can go wrong (illustrative): the assistant reads raised charms as glitter, matches the photo to Tier 1, and the client is booked for 60 minutes instead of 120. The technician either runs late or rebooks the client, whose first impression is of a salon that got it wrong. The fix was a rule, not a better model: any photo-based request goes to "technician review", and the slot is held but not confirmed until one of them has looked. The sibling tutorial on booking gel, acrylic and fills with an AI receptionist covers the standard service lengths.
A week's time log: 32 desk hours, 8 handed back
Take a salon with six technicians and a receptionist on 32 paid hours a week. The receptionist logs the week in half-hour blocks. The right-hand columns are the owner's assumptions about how much AI could take on; every figure is illustrative.
| Duty | Hours a week | Share AI handles | Hours moved |
|---|---|---|---|
| Phone and message enquiries | 9 | 60% | 5.4 |
| Booking, moving and cancelling | 4 | 40% | 1.6 |
| Reminders, confirmations, no-show chasing | 2.5 | 80% | 2.0 |
| Review and social replies | 2 | 50% | 1.0 |
| Check-in and greeting | 6 | 0% | 0 |
| Payments and retail | 5 | 0% | 0 |
| Complaints and problem-solving | 1.5 | 0% | 0 |
| Towels, restocking, hygiene support | 1.5 | 0% | 0 |
| End-of-day admin | 0.5 | 0% | 0 |
| Total | 32 | 10.0 |
Ten hours move, but someone has to read the transcripts and escalations. Twenty minutes a day over a six-day week is two hours, which leaves 8 hours of net saving, a quarter of the role. The sum for the tool: the call-and-message tier of an add-on like Fresha's costs $99.95 a month including 200 call minutes. If the salon takes 120 calls a month at two minutes each, that is 240 minutes, so 40 extra minutes at $0.60, or $24, and a total of $123.95. Eight hours a week is about 35 hours a month, so the tool costs roughly $3.60 for each hour freed.
That comparison only means something if the freed hours go somewhere. Say the desk costs $18 an hour including payroll costs (use your own figure). Cutting paid hours saves real money; keeping them and moving the time to rebooking calls, aftercare messages and retail may earn more, but it is a different decision. Fresha's pricing page showed the calls tier as "coming soon" in at least one market in October 2026, so check that it is offered where you operate before you cost it. The wider comparison with hiring is in the AI receptionist versus front-desk hire tutorial listed at the end.
Turning the split into instructions the AI follows
Once the table is agreed, have an assistant turn it into the instruction sheet you will paste into the booking app's AI settings:
I own a nail salon with [number] technicians. Below are my duty split and my knowledge base. Write the instruction sheet for the AI assistant in our booking software.
Sections: (1) what it may answer from the knowledge base; (2) what it may book and under which conditions; (3) topics it must never decide (money, health, complaints, exceptions) and exactly what it says instead; (4) handover triggers, each with the information it must pass on; (5) the opening line that says it is an AI assistant and how to reach a person.
Use only the facts I give you. Do not invent policies, prices or discounts. If something is missing, write ASK.
DUTY SPLIT: [paste the table]
KNOWLEDGE BASE: [paste hours, prices, policies]Two things went wrong in the example run (illustrative). The draft told the assistant to "offer a free fix to calm an upset client", which is a decision the owner never delegated, so it came out. And it wrote that "our redo policy covers 7 days", a number nobody had supplied; the second draft read "ASK: redo policy period", which is the right behaviour, and the owner then filled in the real figure. Treat every ASK as a to-do for you, not a gap for the AI to fill.
Rewriting the receptionist's role around the split
Once the AI takes the repeat questions, the receptionist's job changes shape. The person spends less time on the phone and more on the door, the till, the queue and the handover inbox. A role summary that reflects that (illustrative): greets every client within half a minute and checks them in; runs the walk-in queue; handles payments, gift cards and retail; reads the AI's transcripts and escalations for 20 minutes a day and replies to anything handed over; updates the knowledge base when prices, hours or policies change; supports the daily hygiene log and signs it.
Two of those duties are new: reading the AI's transcripts and owning the knowledge base. Someone has to own it, because a stale price list makes the AI wrong at speed. If you are writing the advert, the candidate-facing version is a different task; the hair salon receptionist job description tutorial covers drafting one with AI, and you can paste the role summary above into it for a nail salon.
A four-week trial with a stop rule
Do not switch everything on at once. In week one, if your tool offers a draft or review mode, the AI drafts replies and a person sends them, which shows you what it would have said; if it does not, start with the week-two setting. In week two it answers only hours, prices and policies. In week three it books standard services. In week four you review the numbers: how many conversations it handled, how many it handed over, how many wrong answers you found, how many clients asked for a person, and the hours you spent reading.
Add a stop rule before you start: any wrong price, any invented policy, or any promise of a refund pauses that topic until the instructions change. Tell clients they are speaking to an AI. If you have customers in the EU, the AI Act's transparency duty for chatbots has applied since 2 August 2026, and elsewhere it is still good practice, because clients dislike finding out halfway through a conversation. If you answer on WhatsApp, the platform's business messaging policy asks for prompt, clear and direct escalation paths when you use automation, which is exactly what your handover list provides.
Questions owners ask about AI on the front desk
Can the AI take card details or deposits over the phone?
Keep card numbers out of AI conversations. Let the assistant send the deposit or payment link your booking software already generates, so the client pays on the booking app's own payment page, and keep refunds, discounts and disputed charges with a person. If a caller starts reading out a card number, the instruction sheet should tell the assistant to stop them and send the link instead.
How do I stop the AI booking a technician who is off?
Block the time in the booking software's calendar. Booking assistants book against the availability the software shows, so a holiday, a training morning or a late start that lives only in a group chat is invisible to them. Make blocking the time part of approving any leave request, and check the next fortnight's calendar every Monday.
What should I tell my receptionist before switching it on?
Show them the duty split and the time log before anything changes, and ask where they think the AI will get things wrong; they know the awkward calls better than anyone. Be clear whether hours, duties or both will change. If you plan to cut contracted hours or change the role, check the employment contract and take advice first, because that is a contract change, not a software setting.
What if the AI line goes down on a busy Saturday?
Decide the fallback before launch: where calls ring if the AI line fails, who watches the booking app's inbox, and how to switch the AI off, which for a forwarded number usually means turning off call forwarding on your phone line. Run the fallback once during the four-week trial, so the first time anyone uses it is not the morning it matters.
Further reads
- AI Agent for a Beauty Salon: Treatment Questions It Must Hand to Staff — Treatment questions an AI agent must pass to staff.
- Best AI Receptionists for Salons, Compared by Price — Compare receptionist tools by price before you pay.
- AI Receptionist vs Front-Desk Hire: The Real Cost for a Salon — Work out the real cost against hiring a front-desk person.
- How to Write Call Scripts and Escalation Rules for an AI Receptionist — Longer call scripts and escalation rules to adapt.
- How to Measure an AI Receptionist's Return in the First 90 Days — Measure what the trial actually saved.
- What to Do When an AI Receptionist Gets a Booking Wrong — Fix a wrong booking without losing the client.
- Salon Retail Sales: Write Aftercare Product Recommendations With AI — Use freed desk time for aftercare and retail messages.
- Nail Salon Website Built With AI: Pages, Prices and Booking Button — A weekend plan for a nail salon website: pick a builder, prompt it with real facts, lay out prices and add-ons, wire the booking link and get photo consent.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Fresha's pricing and AI Concierge feature pages, Vagaro's pricing page and help centre, and Booksy's AI Receptionist (beta) page, read in October 2026; vendor list prices change, so check the live pages.