The migration tax nobody puts in the AI receptionist pitch

Reva’s Vagaro already knows the colour formulas, the gift-card balances, the booth-renter split, and which 90-minute room cannot overlap. A vendor told her to switch POS anyway. The new logo had AI in the screenshot. The phone, they said, would finally answer once checkout lived in their calendar.

She would have spent the next two Saturdays retraining Colette on a new till, reissuing cards, and watching notes on Noor’s clients disappear into an import that almost worked. The 3:30 hole would still have been a hole. The line would still have rung while every tech was in a service. Switching POS to get a smarter phone line is the wrong trade.

Keep Vagaro, Square, Booksy, Zenoti, Mindbody, or OpenTable if the calendar and the till already work. You do not need a new processor to cover the unanswered ring. You need something that sits beside the book you already trust.

You already paid for a calendar

The till already does the hard part you paid for. Appointments with duration and staff. Checkout, tips, gift cards, packages, memberships. Client records and the card on file. Online booking for people who will use a widget. Reminders, deposits, cancellation rules. Payroll and the report you already send to the accountant. A waitlist button, in one of four flavours, depending on the vendor.

Staff already know the screens. That stack is expensive to move even when the new vendor is polite about imports. The pitch hides the trade. You are not buying a phone line. You are buying a new calendar, a new till, a new payroll map, and a new place for every note Colette wrote on a formula. You retrain Saturday. You tell the desk the checkout they trust is gone. The missed call is still a missed call.

Zenoti published in 2026 that 37 percent of calls to salons and spas go unanswered, and that 82 percent of those missed calls happen during business hours. Vendor data from a company that also sells an AI receptionist. Treat it as a named observation. It still matches the floor. The line rings while Noor is in a 90. Pat cannot leave the walk-in.

Moneypenny’s small-business call report found that 69 percent of callers who hit voicemail decline to leave a message. A message is not a book. A callback at 6pm is not a 2pm gel combo. Booksy’s features page says 32 percent of bookings on their network are made after hours. Online booking can take some of that. It cannot take a voice call. It cannot see that Colette is the only one who can take the fill until 4:10.

The book is not the gap. The work around the book is the gap.

What the till finishes on a Saturday

Name the left-hand column before anyone talks about AI. This is what you keep.

Vagaro already runs the calendar, the marketplace listing, checkout, waitlist modes, and cancellation rules. Square already runs Appointments, the Square till, reminders, and Square Assistant texts that confirm, cancel, and reschedule. Booksy already runs the marketplace book, deposits, self-book, and a waitlist SMS that does not auto-hold the chair. Zenoti already runs rooms, kiosk, memberships, and the spa ledger. Mindbody already runs classes, appointments, and membership cadence. OpenTable already runs the table, the reservation, and the restaurant waitlist.

None of those left-hand cells are a reason to migrate. If Square Assistant already confirms and reschedules the way your clients text, you do not rip Square out to get a different bot. You ask whether after-hours voice and the empty slot still die. The add-on versus the dedicated layer is a separate cut at Vagaro AI vs a dedicated AI receptionist.

Payments stay with the processor you already use. Gift cards do not get reissued because you wanted the phone answered. If you are mid-migration and the only thing you wanted was a voice line, you are paying the tax for the wrong layer.

What still sits around the book

The right-hand column is the work that dies while the till is doing its job.

The call while every tech is in a service. After-hours voice that has to book the right person or room, not leave a voicemail. The cancel that never hits a waitlist because nobody looked up from a fill. A waitlist fill that reads duration, texts once, and will not stack two bodies in one chair or one room. Two languages on the same thread. A stop when the request is a VIP or a dispute. Who is actually open, live, while Saturday is moving. The review ask in the language the client already used. The next visit with the same provider, asked before they walk out.

Online booking can exist and the phone will still ring. A waitlist button can exist and the 3:30 can still go dark because nobody is free to finish the book. That is not a POS failure. That is a gap beside the POS.

A chatbot replies from a tree and never reaches checkout. An answering service takes a name and cannot see that colour cannot overlap. The longer split is AI receptionist vs chatbot for bookings. A book-aware agent has a job, tools on the calendar you already run, and a stop when a person is required. Checkout stays. The person at the desk stays.

The migration tax nobody puts in the pitch

The all-in-one sentence is always the same. Switch, and the phone will finally answer. What it does not say is the two-week Saturday.

Colette has muscle memory on Vagaro checkout. A new till makes her slow, which makes the line behind her longer, which makes the walk-in leave, which is the problem you were trying to solve with a smarter phone. Gift cards have to be reissued or mapped. Packages have a remaining balance that will not survive a sloppy import. Colour formulas live in notes that do not have a field in the new system. Booth renters have a split the accountant already understands. Memberships have a cadence that will bill the wrong day if the import is off by one.

You also lose the week. Every hour spent in training is an hour the floor is worse than it was on the old till. The missed-call problem does not pause for the project plan. If the only job you wanted was a voice that can still book against the live calendar, you can have that job without moving payroll.

Growing is more rings while more techs are in service. It is not a reason to switch POS. The fit question for that buy is AI receptionist for salons on Vagaro. Multi-location is the same agents on separate books, one owner view, hours and menus still local. If you have twenty doors and a POS that is already the wrong POS, do the POS project. Do not hide it inside an agent pitch.

Score the phone, not the logo

Do not start with a winner. Run last Tuesday against the product, whether it is a POS add-on, a chatbot, an answering service, or an agent platform.

It has to book the person who is actually open. Put a gel combo and a 90-minute treatment against the real book. Duration, chair, and room have to match. Inventing a time, double-booking colour, or stacking two bodies in one room is a fail.

It has to stop. A VIP hold and a refund thread leave the agent and land on a person. Talking past the desk, or offering Mrs. Cho’s hold to a walk-in, is a fail.

It has to sit beside checkout. Payment still hits the processor you already use. Payroll does not move. A new till, new gift cards, and a new booth-renter split means you are not buying a phone line.

It has to keep the language the client already used. Spanish SMS books the same way English voice does. An English-only tree on a bilingual desk is a fail.

It has to fill an empty slot once, for the duration that is actually free, without blasting the whole list. It has to be watchable: job, tools, result, or handoff, not a dashboard of conversations. It must not be sold as a firing plan. Pat still owns VIP, disputes, and the person standing at the desk.

Most “AI receptionist” add-ons fail the stop, the two-language desk, or the live chair. Most answering services fail duration. Most chatbots fail the book. The add-on inside a POS can still be the right buy if it passes those tests on your book. If it does not, you have a gap beside the POS. You do not have a reason to migrate payroll.

The demo that proves this is thirty minutes against your book, not a generic deck. What to test on an AI receptionist demo is the agenda: missed call, empty slot, language, stop, and a visit that exists in the POS when you refresh. If the visit only exists in the vendor’s window, you watched theatre.

When you actually do need a new POS

Say it out loud if the till is the problem. Cards on file that will not settle. Rooms the software cannot see. Payroll that cannot split a booth renter. A Saturday that still needs a paper diary next to the screen. Combos the calendar cannot price. That is POS work. Agents will not fix it.

A spa with rooms is a duration problem. If your POS cannot see the room, fix the POS. If it can, and the line still dies during a 90, staff the line. A barber line has to keep moving. Walk-ups will not hold. An agent that promises a chair that is not open will burn the floor faster than voicemail, and a new till will not save it.

A one-chair shop that already answers every ring, including after close, does not need this layer at all. If the tape is empty and the booking page already works, extra software is extra software, whether it is a new POS or an agent.

LLDTEK is an agent platform. Front desk, recovery, rebooking, reputation, and floor sit beside Vagaro, Square, Booksy, Zenoti, Mindbody, OpenTable, or the book you already click. They do not replace checkout. They do not replace the person at the desk. Salons were the first floor we staffed. They are not the product. Clinics, tables, field crews, and studios run the same agents against a different book.

You do not start with five agents. Front desk first, on the line you already publish. The six-chair sequence is How to add an AI receptionist to a six-chair salon. Watch a Saturday before you call it live. There is no public price list. Numbers only if the floor is a match.