By the editors airassvet.ru. The guide was prepared on September 29, 2026. Technical details were checked against the YCLIENTS REST API documentation, and the legal provisions were checked against the legal texts in ConsultantPlus as of the same date. The legal sections are for informational purposes only and do not replace advice from a lawyer. The editors have no commercial relationship with YCLIENTS or other services mentioned, and there are no affiliate links.
Short answer. An AI receptionist for a beauty salon is an AI agent in a messenger or on a website that talks with the client, checks available time slots for stylists in the booking system, and creates or reschedules appointments on its own. In YCLIENTS, the API includes online booking methods for this: available dates, services, stylists, and sessions, booking parameter validation, and appointment creation and rescheduling. AI works best for standard bookings, reminders, and answers to common questions. Complaints, complex medical questions in aesthetics, and promotional outreach require a human and legal review.
Below is which scenarios are worth giving to AI, how integration with YCLIENTS works, where the legal boundaries are, and how to launch a pilot without risking your client base.
Contents
- What an AI receptionist is and how it differs from a chatbot
- Which receptionist tasks can be handed off to AI
- How an AI receptionist books a client through YCLIENTS
- Channels: website, Telegram, MAX, phone
- What to keep with a live receptionist
- Legal limits: personal data and outreach
- Ready-made service or custom development
- How to launch a pilot in four steps
- Pre-launch checklist
- Frequently asked questions
What an AI receptionist is and how it differs from a chatbot
A classic chatbot guides the client with buttons: choose a service, choose a stylist, choose a time. If the client writes in free text — “can I come in Saturday after lunch to Lena for coloring?” — such a bot usually gets lost.
An AI receptionist is an AI agent based on a language model that understands the client’s free text, queries the booking system for available times on its own, and performs an action: creates, reschedules, or cancels an appointment. The difference from a bot is not in “smart answers,” but in the fact that the agent is connected to the salon’s data and can act, not just talk.
This leads to the main requirement: an AI receptionist is only as useful as the quality of its connection to the booking system and the salon knowledge base — pricing, service descriptions, prep instructions, and cancellation policies.
Which receptionist tasks can be handed off to AI
Below is the task breakdown we recommend for a launch. This is an editorial risk-based assessment, not a measurement result.
| Task | Can be handed to AI | Condition |
|---|---|---|
| Booking a standard service with a specific stylist or any available stylist | Yes | Integration with the booking system and an up-to-date service list |
| Rescheduling and canceling an appointment at the client’s request | Yes | Cancellation rules are documented in the knowledge base |
| Visit reminders and confirmation | Yes | This is a service message about an already created booking |
| Answers about pricing, duration, and procedure preparation | Yes | Answers only from the approved salon knowledge base |
| Matching a stylist and time to the client’s preferences | Yes, with caveats | The agent suggests options, and the client chooses |
| Promotions and “come back to us” offers | Only after legal review | May be considered advertising — see the section on outreach |
| Questions about contraindications in aesthetics | No, hand off to a specialist | Medical information and the cost of an error |
| Complaints, conflicts, refunds | No, hand off to a manager | These require a person’s authority and empathy |
Practical principle: AI handles repetitive actions with clear rules, while anything that requires authority, medical qualification, or a decision made at your own risk stays with people.
How an AI receptionist books a client through YCLIENTS
In the “Online Booking” section of the YCLIENTS REST API documentation there are all the steps the agent needs. Simplified, the dialogue “book me for a manicure on Saturday” turns into this call sequence:
- Clarify the service. The method for retrieving the list of bookable services lets the agent match the client’s words to a service in the branch catalog.
- Find a stylist. The method for listing staff available for booking. In the booking parameters, you do not have to choose a stylist: according to the documentation, the specialist ID is passed as 0 if any stylist is selected.
- Find the date and time. The
book_datesmethod returns the days when a staff member or branch has available sessions, and the method for bookable sessions returns the specific time slots. - Check the booking. The booking parameter validation method returns an empty response with status code 201 if the appointment can be created. If the selected time is already taken, the documentation says a 422 response is returned — the agent should offer the client another time instead of double-booking it.
- Create the booking. The
book_recordmethod accepts the client’s phone number, name, email, and booking parameters. There is a separate method for rescheduling an appointment.
The documentation also describes sending an SMS code to confirm the phone number and YCLIENTS webhooks for app events in marketplace applications. The latter is useful if the agent needs to learn about booking changes made manually by an administrator.
Two technical details are worth remembering. First, the pre-creation check is not a formality: while the client is thinking, another client or an administrator may take the slot. Second, in the book_record method, email is marked as a required field; if the client does not provide one in the messenger, the behavior should be agreed in advance with the branch settings and tested. API access is granted through YCLIENTS — the process is described in the YCLIENTS API access guide.
If the salon uses a different booking system, the logic is the same: the agent needs methods to retrieve available times, validate, and create bookings. No API means no full AI receptionist — only a bot consultant.
Channels: website, Telegram, MAX, phone
An AI administrator can be connected to multiple channels at the same time, while the booking logic and knowledge base remain shared.
- Website and widget. The easiest way to start: the client has already come with the intent to book an appointment.
- Telegram and MAX. They’re convenient for reminders and follow-up messages after a visit. We covered connecting an assistant to different messengers in detail in the article AI assistant for WhatsApp, VK, Telegram, and MAX.
- Phone. Voice AI handles missed calls during peak hours, but it requires separate setup for speech recognition and call flows. See our breakdown of voice AI for a call center.
For a salon with one front-desk administrator, a sensible rollout order is website and one messenger first, then phone: that makes it faster to see booking quality and easier to fix mistakes. This is our risk-reduction recommendation, not a technology requirement.
What to keep with a human administrator
The AI administrator should know when to hand the conversation off to a person. Minimum handoff triggers:
- the client is unhappy, complains, or asks to “talk to a person”;
- a question about contraindications, allergies, pregnancy, or medications before a procedure;
- a non-standard service or package that isn’t in the catalog;
- a dispute over price, discounts, refunds, or service quality;
- the agent twice failed to understand the request or got an error from the booking system.
The handoff should be visible: the client gets a message that the administrator will take over, and the administrator sees the full conversation and the data already collected. A silent “pause” with no reply is worse than an honest handoff.
Legal limits: personal data and marketing messages
This section is an informational overview of the rules, not legal advice.
Personal data. A client’s name and phone number are personal data, and their processing is regulated by Federal Law 152-FZ. Also important is Article 10 of the personal data law: health information falls under special categories, and processing it is generally allowed, for example, with the subject’s written consent or by a person professionally engaged in medical practice who is required to keep patient confidentiality. For a salon offering cosmetic services, this means the agent should not collect health information in chat, and such questions should be passed to a specialist. For more on sending client data to external AI services, see the article can client data be sent to AI.
Promotional messages. Under Part 1 of Article 18 of the Advertising Law , advertising distribution through electronic communication networks is allowed only with the subscriber’s prior consent, and the advertiser must be able to prove that consent; upon the recipient’s request, the mailing must be stopped immediately. Part 2 of the same article prohibits advertising distributed using number selection or dialing devices without human involvement — automated dialing and automated mass messaging.
In practice, this separates messages into two types:
- Service — appointment confirmations and reminders for an appointment the client booked themselves. This is not an offer of new services.
- Promotional — promotions, discounts, “you haven’t been here in a while, come by.” These are safer to send only to clients with recorded consent to receive ads and an unsubscribe option, and the automated sending format should be approved by a lawyer.
We covered the penalties for such mailings in the article penalties for AI spam messaging.
Off-the-shelf service or custom development
There are ready-made AI receptionists for salons and agent builders on the market. The choice depends not on how “smart” the model is, but on how well the solution fits the salon’s workflows.
| Criterion | Ready-made service | Custom development or salon-specific customization |
|---|---|---|
| Launch speed | Faster: standard workflows are already built in | Requires a design and testing phase |
| Integration with the booking system | Only with supported systems | Any system with an API, including your own |
| Non-standard rules | Limited by the service’s settings | Any rules and checks can be defined |
| Where data is processed | Determined by the provider | You can choose the model and hosting location |
| Vendor dependence | Higher | Lower, but there are maintenance costs |
If you have one location, YCLIENTS, and standard services, it makes sense to try a ready-made service first; if you have a chain, your own booking system, or strict data requirements, custom development can pay off the time spent on planning. This is an editorial heuristic; compare specific offers using the same set of test conversations. We broke down how to prepare such a task in the article technical requirements for an AI agent.
How to launch a pilot in four steps
A pilot is needed to check booking quality with your clients before letting AI handle the main flow. A process you can repeat in any salon:
- Record baseline metrics for 2–4 weeks. How many inquiries come in through each channel, how many end in a booking, how many are left unanswered outside business hours, and how much time the administrator spends on chat. Without this baseline, there’s nothing to compare the pilot results against.
- Build a test set from 50–100 real inquiries. Use past conversations (anonymized) and include difficult cases: rescheduling, “any stylist,” unavailable time, questions about contraindications, complaints.
- Run the agent on the test set before launch. For each conversation, note whether the service was understood correctly, whether a real open time was offered, and whether the handoff to a person worked where it should. Fix errors in the knowledge base and rules, then run the set again.
- Launch on one channel with monitoring. For the first few weeks, the administrator reviews all bookings created by the agent. Compare them with the baseline from step 1 and expand to more channels only when booking errors are rare and understandable.
AI agent performance metrics are discussed in more detail in our article on AI agent metrics.
Pre-launch checklist
- You have API access to the booking system, and the methods for checking availability and creating appointments have been tested.
- The service catalog, durations, and prices are current and match the booking system.
- The knowledge base covers preparation for procedures, cancellation rules, and late arrival policies.
- The triggers for handing the conversation over to a human and who is responsible at what time are specified.
- The agent does not collect health information in chat.
- Service messages and promotional messages are separated; promotional messages require verified consent and an opt-out.
- It is defined where customer data is processed and which model is used.
- Baseline metrics and a test set of conversations have been recorded.
- An employee has been assigned to review the agent’s records during the first weeks.
Frequently Asked Questions
What can an AI receptionist do for a beauty salon?
It understands the client's free-form text, finds available time slots for stylists in the booking system, creates, reschedules, and cancels appointments, sends visit reminders, and answers questions about prices, duration, and preparation for procedures based on the salon's knowledge base. Complaints and medical questions must be handed over to a human.
Can an AI receptionist be connected to YCLIENTS?
Yes, if API access is available. The YCLIENTS REST API documentation includes online booking methods: lists of available services, staff, dates, and sessions, booking parameter checks, appointment creation and rescheduling, as well as event webhooks for marketplace apps.
Can an AI receptionist be used to send promotional campaigns?
Only with caution. Under Article 18 of the Advertising Law, advertising sent through telecommunications networks is permitted with the subscriber's prior consent, and automated advertising mailings without human involvement are prohibited. Appointment reminders are service messages; promotions and return offers are best reviewed with a lawyer.
Will AI replace a live receptionist?
Not completely. AI handles routine bookings, rescheduling, reminders, and common questions, while complaints, unusual services, disputes, and medical questions remain with people. The receptionist gets more time to spend with clients in the salon.
Where should you start when implementing an AI receptionist?
With baseline metrics and a test set of real inquiries. Then the agent is tested on this set before launch, rolled out on one channel, and the receptionist checks all created bookings during the first few weeks.
How the airassvet.ru team helps implement an AI receptionist in a salon
If an off-the-shelf service does not fit your processes, booking system, or data requirements, the team airassvet.ru can:
- analyze your chats and calls, document booking scenarios and human handoff triggers, and build a test set of conversations;
- build a salon knowledge base—services, procedure preparation, cancellation rules—so the agent answers only from it;
- develop an AI agent for your website and messaging apps, integrate it with the booking system and CRM, test it on a conversation set, and train your team to use it.
The first step is to choose one process, such as website booking, document how it works now, what data and constraints it has, and what criterion will define a successful pilot. If the process is selected, discuss the project.
Conclusion
An AI receptionist for a beauty salon is not a button-based chatbot, but an agent connected to the booking system and knowledge base. YCLIENTS has all the API methods needed for this: available times, booking validation with protection against occupied slots, creation, and rescheduling. AI should handle routine bookings, rescheduling, reminders, and common questions, while complaints, medical questions in cosmetology, and promotional mailings should be left to people and legal review. Start with baseline metrics, a test set of conversations, and one channel—that way you can see the real impact, not just promises.