In the work of a beauty or health specialist, there’s one constant: messages. Clients write at any time — at seven in the morning, late at night, on weekends. “Do you have anything on Wednesday?”, “Can I book with you next week?”, “Remind me of the address”, “What if I move it to Friday?”.
Answering all of this manually means being available constantly. Not answering means losing clients who simply go to whoever replied faster.
This is exactly the task an AI agent in Telegram was set up for. Here’s what it looks like from the inside and what changed in day-to-day work.
What’s inside:
- The specific task the agent was solving
- What the booking process through the agent looks like — step by step
- How reminders work
- What the agent handles — and what stays with the specialist
- What went wrong on the first launch — and how it was fixed
The task: book clients without manual involvement
Before the agent, bookings looked like this: a client writes in Telegram, asks about availability, gets a reply (sometimes after several hours), then clarifies details, then confirms — and only after all that is the booking logged manually in a spreadsheet or notes.
One booking takes an average of 10-15 minutes of back-and-forth. With 20-30 bookings a week, that’s several hours spent purely on communication.
The task for the agent was simple: accept a booking request, check the calendar, confirm — all without the specialist’s involvement.
How the agent works: step by step
Here’s what a typical client conversation with the agent looks like:
Client: Good afternoon, I’d like to book a haircut
Agent: Good afternoon! What date and time would work for you?
Client: Can I come on Thursday afternoon?
Agent: Thursday has the following slots available: 2:00 PM, 3:30 PM and 5:00 PM. Which time suits you?
Client: 3:30 PM
Agent: Perfect! You’re booked for Thursday at 3:30 PM. I’ll send you a reminder the day before and an hour before your visit. If you need to reschedule — just write to this chat.
The whole conversation: under two minutes. The client has their booking, the agent has logged it in the calendar, the specialist has received a notification.
For this, the agent is connected to the calendar via an integration with Calendly or Google Calendar. When the client picks a time — the agent checks real available slots and blocks the chosen one.
How reminders work
Reminders turned out to be just as important as the booking itself.
The agent sends the client two automatic messages:
24 hours before: “Reminder: your appointment is tomorrow at 3:30 PM. Address: [address]. If you need to reschedule — please write to me by 8:00 PM today.”
2 hours before: “We’re looking forward to seeing you today at 3:30 PM. If anything has changed — just let me know.”
Before the agent, reminders weren’t sent at all — no time for it. The number of clients who simply didn’t show up without notice was 2-3 per week. After — practically zero.
Every missed appointment is lost time and lost income. The agent solved this problem completely.
What the agent handles — and what stays with the specialist
The agent reliably handles predictable tasks:
✓ Accept a booking request ✓ Show available slots ✓ Confirm and log the booking ✓ Send appointment reminder ✓ Answer standard questions: price, address, treatment duration ✓ Reschedule a booking on the client’s request
The specialist keeps:
✗ Unusual situations — a client writes something complex or emotional ✗ Consultations about treatments — the agent doesn’t replace expert conversation ✗ Conflict situations — an unhappy client should speak to a real person ✗ New clients with non-standard requests — better to handle the first time manually
A well-configured agent knows when to hand off a conversation. A trigger is built into the settings: if a client writes “want to talk”, “have a question”, or uses emotional language — the agent replies “I’ll pass your request on, someone will be in touch shortly” and notifies the specialist.
What went wrong on the first launch
The first version of the agent was too rigid. Clients could only choose from pre-set options — buttons. Those who wrote in free text (“do you have anything Friday after 6?”) got the reply “please choose one of the options provided”, which was frustrating.
The second problem was unusual requests on first contact: “what’s better — a haircut or colour?” The agent tried to answer but did it poorly.
After the fix:
- The agent now understands free text with dates and times
- Consultation questions now get the response: “Better to discuss this in person — write to us and you’ll get a reply within an hour”
- All requests the agent couldn’t process go into a separate list for manual reply
The rework took about two weeks of work with real conversations. After that — stable.
What changed in day-to-day work
Before: messaging took 2-3 hours a day, several clients a week didn’t show up without notice, some requests got lost in the evenings and on weekends.
After the agent: booking messages — about 20 minutes a day (only unusual cases), no-shows without notice — rare exceptions, requests at any time are handled immediately.
This doesn’t mean the agent was perfect from day one. It means that after a few weeks of configuration, the system runs by itself — freeing up time for actual work with clients, not administration.
How it’s built technically (simply)
- Telegram bot — the front end the client talks to
- Agent logic (Make.com or n8n) — the brain that understands the request and decides what to do
- Calendar (Google Calendar or Calendly) — the source of availability
- Notifications — agent sends confirmation to the client and notification to the specialist
More on how AI agents for small business work in general — there’s a dedicated article on the principles.
Related reading
- Automate client work without hiring staff — what else can be automated beyond bookings
- Instagram chatbots: what actually works — a look at typical expectations vs reality
- Small business automation: where to start — if you’re just starting to explore the topic
Frequently Asked Questions
What is an AI agent for client bookings? A Telegram bot with extended logic: it understands free text, checks the calendar in real time, confirms bookings and automatically sends reminders — unlike a simple bot, it can hold a real conversation.
Do you need technical knowledge to set it up? For basic configuration — no. Tools like Make.com allow you to build the logic without coding. For more complex scenarios, a specialist’s help is needed.
What to do if the agent doesn’t understand the client? Set up a trigger: if the agent doesn’t recognise the request — automatically notify the specialist. The client receives: “I’ll pass your request on, someone will be in touch shortly.”
How long does setup take? Basic version — 3-5 hours. Refinement based on real conversations — another 1-2 weeks of observation.
Does the agent handle rescheduling? Yes, if that’s in the scenario. Client writes “I want to reschedule” — agent offers available slots and updates the booking in the calendar.
Can the agent be set up without Calendly? Yes. The agent can be connected to Google Sheets, Notion, or any other tool where the schedule is stored. Calendly is convenient because it has a ready-made API and blocks booked slots automatically.
Leave a Reply