How an AI Bot Replaced Two Call-Centre Operators in a Medical Clinic

Automating patient bookings in a clinic isn't about "getting rid of humans." It's about not making your admin answer "what time do you open?" for the eighth time today so they can focus on patients who actually need help. In this article we walk through a typical scenario: how a 21-day deployment of an AI chatbot on amoCRM Salesbot + WhatsApp Business API unfolds for mid-sized clinics with 8–15 doctors.

About this format. Below is a typical scenario: which tasks get closed, which stack we use, which effect ranges we see on projects in this niche. It is not a report on a specific clinic (we publish one named case with the client's written permission — see Cases on the homepage). Figures are characteristic ranges from similar projects, not a guarantee for any specific client.

Where the automation idea came from

Before this kind of project we typically audit the contact centre — one month of inbound, around 3,000–5,000 messages and calls. The categorisation usually shows: 60–70% of inbound is 5–8 types of identical questions. Doctor schedules, appointment cost, address, parking, opening hours, available slots, rescheduling. Every one of these flows can be formalised and handled by a bot, no human needed.

The remaining 32% are either complex clinical questions (need a doctor or senior admin) or complaints (need a human). The bot leaves these alone — it routes them to an operator with a tag describing the topic.

What we automated through amoCRM Salesbot + GPT

• Specialist booking — the bot asks what kind of help is needed, offers 3 nearest slots, confirms in WhatsApp and auto-creates a deal in amoCRM
• Reschedule and cancel — fully autonomous, no operator. The freed-up slot returns to the public pool in 30 seconds
• FAQ: doctors, prices, address, hours, parking, insurance — answers pulled from a knowledge base the clinic owns and updates itself
• Appointment reminders at 24h and 2h before — no-show rate dropped from 18% to 4% in the first month
• Satisfaction survey after the visit — automatically in WhatsApp or Telegram, aggregated reporting per doctor in amoCRM

Technical implementation without developers

The bot runs on amoCRM Salesbot — the built-in visual scenario builder. For complex queries and fuzzy inputs ("something for back pain, which doctor do I need?") we wire in a modern LLM (e.g. GPT-5 or Claude) through a webhook in Salesbot. The clinic's knowledge base is a single document maintained by the senior admin — no developer involvement.

WhatsApp connection to amoCRM is through the official WhatsApp Business API (not third-party gateways). This matters: unofficial integrations get banned by Meta in waves, and a clinic can lose its main patient communication channel overnight.

No code. No daily dev ops. Everything through the amoCRM interface and WhatsApp Business panel. Full deployment with admin training typically takes 21 days.

Typical results on projects at this scale

• Average response time — from hour-long delays to 1–3 minutes
• Admin workload — typically down 40–60%, freed-up time goes to complex cases
• No-shows — typically from 15–20% to 3–6% within 1–2 months of stabilisation (double reminders)
• Conversion from inquiry to booked patient — typically up by 15–25 percentage points
• Cost of the solution — from €1,500 (AI automation deployment) + from €500/month (basic support)

Payback math on this kind of project: one saved admin headcount (around €1,400/month including ES employer taxes) typically covers deployment plus support in three months. After that — pure savings and higher conversion.

Numbers are ranges from projects in our niche, not a promise to any specific client.