How does a clinic train the intusell AI? A step-by-step guide

A clinic's inbox is always full of the same questions: "How much does a hair transplant cost?", "When do you have an opening for fillers?", "Should I come to the first examination on an empty stomach?", "How many sessions does it take?". Aesthetics, dental, hair transplant, physiotherapy, or dietitian — it doesn't matter; the same twenty questions repeat a hundred times a day. While reception tries to keep up with the phone and WhatsApp, dozens of "price?" comments pile up under a before/after post, and a patient writing in English from abroad waits for hours. Every appointment request answered late goes to another clinic.

intusell takes over this load. But to do that, it first needs to learn your clinic, your services, and your operating routine. This article explains how a clinic trains the intusell AI from scratch, step by step. This is the health chapter of our sector-based "how to train your AI" series, and it is the series' main article.

Quick answer

Why does training matter?

intusell is not an off-the-shelf chatbot; it is a fully autonomous AI assistant that behaves like an experienced receptionist and patient coordinator. A good coordinator does not start on day one without knowing the clinic either: they need to know which services you offer, your appointment routine, and which questions to leave to the physician. Training ensures the AI does two things correctly: give the right information (your services, preparation instructions, working hours) and stop at the right boundary (leave medical matters to the physician, never promise outcomes).

An untrained assistant either gives overly generic answers or guesses on a topic it doesn't know. In healthcare, guessing is dangerous: a wrong piece of preparation information wastes an appointment, and a boundary-crossing "don't worry, it's a sure fix" sentence both misleads the patient and violates regulations. A well-trained assistant stops when it doesn't know and directs the patient to the physician or the live team.

Who is it for?

This guide is for clinics whose inbox is dominated by appointment requests and who answer the same questions over and over again:

• Dental, oral health, and orthodontics clinics
• Aesthetics, medical aesthetics, and dermatology clinics
• Hair transplant centers (often with international patient traffic)
• Appointment-heavy specialties such as physiotherapy, dietitian, ophthalmology, and polyclinics
• Teams receiving dozens of appointment requests a day from Instagram and WhatsApp
• Centers that need to communicate in multiple languages for health tourism

You don't need a technical team. All of the steps below are completed from sections in the panel, without writing code.

The structure of training: 5 core steps + 2 continuous steps

Splitting the training into two groups makes your job easier. The first 5 core steps get the assistant up and running; these steps are the one-time part of the setup. The 2 continuous steps that follow (label review and opening channels) are the ongoing part that you sharpen as you use the assistant. You'll find all of them in order below.

1. Upload the knowledge base

The foundation of training is the knowledge base (RAG). In the panel's Knowledge Base section, you upload all of your clinic's textual information. Each time the AI generates a response, it automatically searches these sources and uses only the information written there.

Supported source types:

Every file you upload is automatically chunked and made searchable with pgvector. When you upload a "pre-implantation preparation" document, the AI finds the right paragraph and responds when a patient asks "Can I smoke before the procedure?".

What a clinic must definitely upload: service and procedure descriptions, pre/post-procedure preparation instructions, working hours, team and physician information, payment/communication policy, and the most frequently asked questions. An important limit: the knowledge base is not for diagnosis-containing text. The AI uses the information here to answer "what do you offer, how do you prepare" questions; it leaves medical questions like "I have this, what should I do?" to the physician.

2. Set up the appointment system and working hours

The knowledge base explains "what you offer"; the appointment system, in turn, lets the assistant give the patient an actual time. In tourism this step was the price catalog; in a clinic its equivalent is the appointment infrastructure. In the panel you define three things:

Once these three are defined, the AI no longer gives times out of thin air; it suggests appointments based on your working hours and full slots. Two appointments at the same time (double-booking) are prevented by the system. When an appointment is created, if Google Calendar is connected, the appointment is also copied one-way to the team's calendar.

We've gathered all the details of the appointment engine — durations, break times, conflict prevention, reminders, and calendar sync — in a separate article: clinic appointment automation, calendar, and reducing no-shows. You can also see the service and appointment capabilities on the solutions page.

3. Set the persona and tone

You taught the AI what to say in the first two steps; now you'll define how it will say it. There are two settings in the panel:

• Assistant name (ai_persona_name): The name it will introduce itself with to the patient. Most clinics choose a reassuring name consistent with their coordination team.
• Tone (ai_tone): For health, the default tone is professional; warm but measured, calm and reassuring. Emoji use is off and sales pressure is kept low.

Health mode is deliberately more cautious than other sectors. Even the greeting message directs the patient to the right place; a typical opening reads like this: "Hello. How can I help you? For medical advice, please consult your doctor." This line brings the clinic's tone and the requirement of regulations together in a single sentence.

A multilingual note: The AI responds in whatever language the patient writes in. For messages arriving from abroad in English, Arabic, or German to hair transplant, dental, and aesthetics clinics, you do nothing extra; the assistant automatically detects the language and responds in that language while preserving the health tone.

4. Define the response rules and health limits

The persona defines "who it is," and the response rules define "how it will behave." In a clinic, this step does two jobs at once: conveying your sales discipline and drawing the health limits. With the AI Manager Chat, you add rules by writing in plain language, just as you would instruct a teammate.

Typical rules for clinics:

• "Before the appointment, ask about the patient's request and suitability, but do not diagnose."
• "When asked about the treatment fee, do not quote a price; say the fee will be finalized after the examination and direct them to an appointment." (Can be changed according to your clinic policy.)
• "Do not guarantee outcomes; do not use phrases like 'sure fix/guaranteed.'"
• "In a medical complaint or emergency, direct the patient to the physician or, if necessary, the emergency service."

The most common mistake in healthcare is misleading the patient and running afoul of regulations with a single boundary-crossing sentence. In intusell, this boundary is not left solely to the rules you write; there is a built-in medical safety layer. Misleading health claims such as "guaranteed treatment," "certain cure," "cures cancer," "heals diabetes," or "Ministry of Health approved" are flagged by pattern scanning and a second model check (LLM judge) before the response reaches the patient. In other words, even if you forget to write a rule, the system tries to protect you from a misleading claim.

If you want to compare two different approaches, you can use the A/B test feature: for example, you can put a more informative tone next to a more directive appointment closing at 50%-50% traffic and measure which one converts to more appointments.

5. Teach from past conversations

This is the step that moves training from "good" to "specific to your clinic." In the panel, you upload audio recordings of your past patient conversations (MP3, MP4, WAV, M4A) and mark each one as Won or Lost — for example, whether the patient came to the appointment and accepted the treatment, or was lost.

The system uses these recordings in two ways:

This way, the AI learns how your best coordinator put the patient at ease and moved someone who said "let me think about it" toward an appointment. The KVKK side is protected: PII (personal data) masking and explicit consent are applied to the uploaded recordings. Because health data is special-category, this protection is even more important in healthcare.

This step is not mandatory, but don't skip it. The knowledge base teaches the AI "what" it knows; past conversations teach "how you coordinate." When the two are combined, the assistant truly resembles an experienced patient coordinator.

6. Correct with label review (continuous step)

The first five steps get the AI up and running. From here on is the continuous part that perfects it over time. Every AI response lands in a label review queue. Here you can do three things: approve, reject, or correct.

The AI learns from these corrections. Say it explained a preparation instruction incompletely or referred to a service by the wrong name; you correct it, and in similar cases it will now use the right wording. Over time, patterns specific to your clinic build up (the procedure names you use, your standard informational sentences, your appointment-directing style).

For the first two weeks, we recommend spending 10-15 minutes a day on this queue. During this period, the correction rate drops quickly, because the AI learns the frequently made mistakes. Review is the "live" part of training: the system gets smarter as it is used.

7. Open the channels (continuous step)

Once training is ready, you put the assistant in front of patients. intusell gathers all channels in a single inbox: Instagram DM, Instagram comments, WhatsApp, Facebook Messenger, Telegram, Web Chat, and email.

Channel-opening methods:

• Meta channels (Instagram, Messenger): connect with one-click OAuth.
• WhatsApp: connects in about 1 minute by scanning a QR code from your phone — no Meta Business approval required. The official Cloud API option is also available.

Thanks to channel gates, you control each channel separately. The Instagram DM gate (instagram_dm_enabled) is separate from the Instagram comments gate (instagram_comments_enabled); for example, you can first automate only DMs and keep the comments under before/after posts manual, then open the comments gate as well when you're ready.

There are working modes for handing off to a human: ai_only (the AI answers everything), human_only (everything goes to the live team), hybrid (the AI normally answers and escalates to the team when needed). Most clinics start with hybrid; the AI handles appointment and information traffic, and conversations requiring a medical decision go to the team. For the details of Instagram and WhatsApp automation: Instagram and WhatsApp appointment automation for clinics.

How long does training take?

A working setup takes half a day:

1. Uploading the first files to the knowledge base: 1-2 hours (shorter if your existing documents are ready)
2. Appointment types and working hours: 20-30 minutes
3. Persona, tone, and first response rules: 30 minutes
4. Connecting channels: 1-5 minutes per channel

But there is no moment of "training is done." Over the first two weeks, as you approve and correct responses in the label review queue, the assistant sharpens to fit your clinic. Uploading past conversation recordings and running A/B tests are also ongoing improvements. Setup is fast; mastery is continuous.

What it isn't

Putting intusell in the right category matters, because the wrong expectation leads to the wrong setup.

• It is not a physician or a diagnostic tool. It does not diagnose, recommend treatment, or give medication dose/recommendations. It directs medical questions to the physician and the clinic.
• It is not a hospital information system (HIS) or an electronic medical record (EMR). It does not keep medical records; it is an appointment, communication, and patient coordination layer.
• It is not a marketing tool that promises outcomes. Phrases like "certain outcome/guaranteed" are both prohibited and constrained so the AI will not produce them.
• It is not an intermediary that replaces the clinic. The one providing the service and issuing the invoice is always the clinic; intusell is not a party to payment.

In short: not a bot that gives canned answers, but a patient assistant that represents your clinic to the extent you train it — while never crossing the medical boundary.

Frequently asked questions

How long does clinic AI training take?

A working setup is completed within half a day: uploading the knowledge base, appointment types and working hours, persona and tone settings, and a few response rules. The real refinement builds up over the first weeks as you approve and correct responses in the label review queue. Training is not one-time; it is continuous.

Does the AI give patients medical advice or a diagnosis?

No. In health mode the AI does not diagnose, recommend treatment, or guarantee outcomes; it directs medical questions to the physician and the clinic. The primary safeguard is the "do not give diagnosis/treatment advice" instruction that health mode gives the AI; in addition, a medical safety layer scans for and flags misleading phrases such as "guaranteed treatment" or "certain cure."

Do I need technical knowledge to train the AI?

No. All training is done from the panel; no code, API key, or developer is required. You upload PDF/Excel files to the knowledge base, enter appointment types and working hours, and write response rules in plain language. You connect channels with one-click OAuth or by scanning a QR code.

Is patient data safe under KVKK?

Yes. Health data is special-category personal data under KVKK and requires explicit consent. intusell is end-to-end encrypted and KVKK-compliant; the sensitive health note field is processed with high sensitivity and manual approval, and the time and type of explicit consent are recorded.

Does the AI quote an appointment price or treatment fee?

It behaves according to your clinic policy and health regulations. It does not market treatment fees on public channels; in a typical setup it defers the fee until after the examination/consultation or directs the patient to the live team. When it is unsure, it does not invent a price.

Does the AI create the appointment on its own?

Yes. The AI recommends the appropriate service and creates the appointment while accounting for working hours and full slots; double-booking the same time is prevented by the system. In cases requiring a medical decision or complex situations, it hands the conversation off to the live team.

Next step

You've trained your assistant; next comes using it in daily operations. The next article in the series explains how to run the intusell you trained on a real workday: incoming appointment requests, calendar, reminders, reducing no-shows, and handing off to the team. Continue straight from there: how a clinic uses intusell.

If you'd like to see it live before starting the setup, use Request a demo and we'll open your panel together in a 20-minute session, or write to hello@intusell.com. For package and quota details, you can check the pricing page, and for other guides, take a look at the all articles list.