AI Receptionist for Salons and Barbershops: How to Build One That Books Real Appointments

Why salons lose bookings to the phone

A stylist mid-color cannot answer the phone. The front desk, if there is one, is checking someone out. So a new client who calls on a Saturday morning gets voicemail, and a person who wants a haircut will usually call the next salon on the map instead of leaving a message.

Here is the illustrative math. Say a five-chair salon misses 10 calls a week during services, and a quarter of those callers are new clients ready to book at an average ticket of $85. That is about two lost new clients a week, roughly $170, or over $8,000 a year before you count repeat visits. Swap in your own numbers: check your phone's missed call log for a week and count how many were unknown numbers.

An AI receptionist picks up those calls, finds a slot and books it. It is not magic and it can go wrong in specific ways, which this guide covers.

What the receptionist has to do on a salon call

A salon call is more structured than most people expect. The caller wants a service, with a particular stylist or any stylist, at a time that fits both of them. The assistant has to know four things.

The service menu, with durations. A men's cut is 30 minutes, a balayage might be three hours with a gap in the middle while color processes. If the assistant books the wrong length, the whole day's schedule is off.

Who does what. Not every stylist does color, cuts curly hair or works Mondays. This has to be data the assistant reads, not something it guesses.

Your rules. Deposits for long services, a 24-hour cancellation policy, a patch test before certain color work, and what to say to someone who wants a service you do not offer.

When to hand off. Bridal parties, corrective color and complaints should go to a human, not the AI.

One builder documented on this site sells a Vapi receptionist for salons and barbershops as a $3,000 package, built with Vapi, Make, GoHighLevel and ElevenLabs (builtwithagents.ai/strategy/salon-ai-receptionist-3000-vapi-make). That price is the builder's own offer and has not been verified against signed deals. It is useful as a signal of what agencies charge, which matters if you are building this for other salons.

Check your booking software first

The assistant is only as good as the calendar it reads. Before choosing any AI tool, find out whether your booking system lets software check availability and create appointments.

Square Appointments has a Bookings API, so it is the easiest to connect to a custom build. Boulevard describes an open API covering appointments, clients, staff and services. Vagaro lists API access as a feature, but I could not confirm the scope. For Fresha I found no official public API, only an unofficial third-party one, which is risky to build a business on. These are findings from secondary sources, so check each vendor's developer documentation before you commit.

If your software has no usable API, use a calendar that does: Google Calendar works, and Cal.com is used in a plumbing build on this site (builtwithagents.ai/strategy/plumbing-ai-receptionist-vapi-make-calcom). The catch is double entry: if clients also book through your main system, two calendars will drift apart. A single source of truth beats a clever assistant.

What it costs

On Vapi's pricing page (as of October 2026) the hosting fee is $0.05 a minute. The other parts are billed at cost: speech-to-text around $0.01 a minute, the AI model between about $0.008 and $0.045, the voice between about $0.015 and $0.024, and Twilio inbound calls at $0.008 a minute. Add it up and a call costs roughly $0.09 to $0.14 a minute.

A salon getting about 100 three-minute calls a month uses around 300 minutes, so $27 to $41 in call costs. Make's free plan includes 1,000 credits a month, which covers a small salon's booking lookups at first; paid plans start at roughly $9 to $11 a month depending on billing, so confirm on Make's pricing page. Vapi's free credits are small ($5), so expect to add a payment method during testing.

Compare that to a part-time receptionist or an answering service charging per call. The software is cheap. The cost is your time to set it up and check it.

Step by step: build it on Vapi and Make

Step 1: Write the service menu as plain data

Open a spreadsheet with one row per service: name, duration in minutes, price, which stylists do it, and any note such as "needs patch test" or "deposit required". Add a second tab for stylist working hours. This sheet is what you will paste into the assistant's instructions and later use to validate bookings.

Step 2: Choose where bookings live

Pick the one calendar that holds appointments. If it is Square Appointments or Boulevard, read the developer docs for the endpoints that list availability and create a booking. Otherwise create a Google Calendar per stylist and use only that for AI bookings, and move the rest of the team to it.

Step 3: Create the assistant in Vapi

Create a new assistant. In the system prompt, say who it is (the salon's front desk assistant), give it the service menu, hours and policies, and state three hard rules: never offer a time that did not come from the calendar tool, never quote a price that is not in the menu, and transfer to a human for anything it cannot answer. Keep the greeting short: "Thanks for calling Lucy's Salon, this is the booking assistant. How can I help?"

Pick a small, fast model for the conversation. Latency matters more than cleverness on a phone call.

Step 4: Connect the calendar through Make

In Make, create a scenario that starts with a webhook from Vapi. Build two paths. The first looks up open slots for a service and date and returns up to three. The second creates the appointment with the caller's name, phone number, service and stylist, and returns a confirmation. In Vapi, add these as tools the assistant can call. Test each path from Make's run-once button before connecting it to a live call.

Step 5: Add the human fallback

Add a transfer tool that forwards the call to the owner or front desk number, with a message that tells the assistant to take a name and number instead if nobody picks up. Then set your carrier's call forwarding to send calls to the assistant's number after a few rings or when the line is busy. The salon number stays the same, and the AI catches only what staff miss.

Step 6: Run 20 scripted test calls

Call it yourself with awkward requests: a service you do not offer, a slot that is full, a name with unusual spelling, a request for a stylist who is off that day, a caller who changes their mind twice. Check each booking against the calendar. Fix the prompt after every failure and run the whole set again.

Step 7: Review transcripts for the first week

Read every transcript for seven days. Look for misheard names, wrong durations and promises you do not want it to make. Send a text confirmation after each booking so a bad capture is caught by the client within minutes.

What goes wrong

The vet clinic receptionist documented on this site reached about 60 percent accuracy in early testing, with trouble capturing names and phone numbers and sometimes inventing appointment slots (builtwithagents.ai/strategy/voice-ai-receptionist-veterinary-clinics). That is a prototype result, but it points at the same failure modes you will hit. The fixes are the ones above: slots only from the calendar tool, a text confirmation, and a person reviewing the first week of calls.

Other problems to expect: callers with strong background noise, clients who say "same as last time" (the assistant does not know), and double bookings if two systems write to the same calendar. Tell regulars the assistant exists so it is not a surprise.

When not to build this

If you run one or two chairs and your clients already book through an online link, you may need nothing more than a voicemail greeting that sends people to the link. If most of your business is high-touch consultations, such as color corrections, a human voice is part of the product. And if you cannot spend a week reviewing calls, wait: an unreviewed assistant booking the wrong service for a month costs more than the missed calls did.

Frequently Asked Questions