العربية, Arabic version of this page WhatsApp us

AutoGreet AI answers your patients on WhatsApp and hands anything medical to a person.

A reception assistant on the clinic’s own WhatsApp Business number. It answers from the facts your clinic enters, takes booking requests, and gives no medical advice. Every conversation shows in a staff inbox, and a person can take over at any time.

Ask for AutoGreet AI See pricing

What it does

Answers from your clinic’s facts only
Opening hours, address and directions, parking, what to bring, languages spoken, the doctors and what they treat, service prices and accepted insurers as guidance, and the questions you add. A question outside the desk gets a polite redirect.
Takes booking requests
A patient asks for a time. By default the assistant notes the request and tells the patient the clinic will confirm, and it never invents a time. A clinic that has attested that the ITIG Clinic calendar is its source of truth can let the assistant book, reschedule, and cancel from the live calendar, with times taken only from that calendar.
Replies in the patient’s language
Arabic, French, English, or Lebanese Arabic written in Latin letters. It mirrors the language and the register the patient used.
Says what it is
The first automated reply in each conversation carries a disclosure line naming the assistant and the clinic. Patients are never left thinking they are talking to a person.
Runs on the clinic’s own number
Your WhatsApp Business number, connected through Meta’s official onboarding on the WhatsApp Cloud API. Meta’s review of a newly connected number takes one to three weeks.
Hands over cleanly
A staff inbox shows every conversation and who holds it. A staff reply takes the conversation over. A patient can be marked always-human. An escalation stays flagged until a person acknowledges it.

Medical messages go to a person, always

Before any model is called, a deterministic check classifies the message. A message that reads as medical, urgent, or an emergency never reaches the model. The patient receives a holding text with the Lebanese Red Cross number, and the conversation is handed to your staff. This cannot be switched off per clinic.

An emergency message is answered even while a staff member holds the conversation, because that is the one thing that must never wait.

In our own test of 48 emergency descriptions in Arabic, Arabizi, French, and English, written blind and none of which the system was tuned against, every one was handed to a person with the 140 instruction, and none of 24 ordinary booking, pricing, and directions messages was flagged. Measured as the worst of three runs on the configuration that ships.

The assistant never states a test result, a value, or an interpretation. If a patient asks whether a result arrived, it can say only whether a document has arrived and when.

What happens to an urgent message

The classification runs on the words of the message before any model does. For an emergency, an urgent message, a complaint, or a patient asking for a person, the model is not called at all.

What AutoGreet AI does with an urgent message A vertical flow. A patient message arrives on the clinic WhatsApp number. A keyword check reads the words before any model runs. The flow then splits in two. An ordinary question goes to the assistant, which drafts a reply, and every claim in the draft is checked against the answers the clinic wrote before it is sent. An urgent message or an emergency takes the other path, where the model is never called at all, a fixed reply gives the Lebanese Red Cross number 140, and the conversation waits for a person at the clinic. A patient writes to the clinicon its own WhatsApp number A keyword check reads thewords, before any model runs An ordinary question Urgent, or an emergency The assistant drafts a replyand every claim in it ischecked against the answersthe clinic wrote The model is nevercalled at all The reply is sent A fixed reply gives theLebanese Red Cross number,140, and the conversationwaits for a person
The reply reaches the patient inside WhatsApp’s 24 hour window. Outside it the conversation is still flagged for the clinic, and the patient receives nothing until the clinic writes first.
What AutoGreet AI does with an urgent message A vertical flow. A patient message arrives on the clinic WhatsApp number. A keyword check reads the words before any model runs. The flow then splits in two. An ordinary question goes to the assistant, which drafts a reply, and every claim in the draft is checked against the answers the clinic wrote before it is sent. An urgent message or an emergency takes the other path, where the model is never called at all, a fixed reply gives the Lebanese Red Cross number 140, and the conversation waits for a person at the clinic. A patient writes tothe clinic on itsown WhatsApp numberA keyword check reads thewords, before any model runsAn ordinary questionThe assistant drafts areply and every claim init is checked against theanswers the clinic wroteThe reply is sentUrgent, or an emergencyThe model is nevercalled at allA fixed reply gives theLebanese Red Cross number,140, and the conversationwaits for a person
The reply reaches the patient inside WhatsApp’s 24 hour window. Outside it the conversation is still flagged for the clinic, and the patient receives nothing until the clinic writes first.

What patients actually receive

These are the product’s own reply templates, quoted as they ship. The clinic’s and the assistant’s names fill the placeholders.

The disclosure line, first automated reply of a conversation

{assistant} is the automated assistant of {clinic} - the clinic team reviews these conversations.

A booking request

I can’t see the live calendar, so I can’t confirm a time myself - but I’ve noted your request and the team will follow up shortly to confirm a slot with you.

A medical question

I’m passing your message to the clinic team because I can’t provide medical advice here. They’ll reply as soon as they can. If this gets worse or starts to feel urgent while you wait, call the Lebanese Red Cross on 140 or go to the nearest hospital.

An emergency, in English and in Lebanese Arabic

This may need urgent medical attention. Do not wait for a reply here. Call the Lebanese Red Cross on 140 now or go to the nearest hospital emergency department.

The same emergency text in Lebanese Arabic

هالحالة ممكن تحتاج عناية طبية طارئة. ما تنطر رد هون. اتصل فوراً بالصليب الأحمر اللبناني عالـ140 أو توجّه لأقرب طوارئ بالمستشفى.

A question outside the desk

That one’s outside my desk - I’m the clinic’s receptionist. Happy to help with appointments, hours, prices, or directions.

After a patient writes STOP

You have been unsubscribed and will not receive further automated messages from {clinic}. Reply START to opt back in.

What the clinic sets, and what it can try

Everything the assistant may say is typed by the clinic: what each doctor treats, the opening hours, the prices, the answers to the questions patients ask. The clinic can put a message in and read the reply before any patient sees one. Conversations the assistant hands over wait in an inbox, with who holds each one.

The screen where a clinic types a patient message and reads what its assistant would answer
Trying the assistant before any patient does. Demo environment with sample data.
The screen where a clinic writes what each of its doctors treats, which is what the assistant is allowed to say
What the assistant may say. Demo environment with sample data.

How a message reaches you

Your clinic connects its own WhatsApp Business number. Nothing passes through a number belonging to ITIG Tech, and there is no SMS anywhere in this path.

How a WhatsApp message reaches the clinic A horizontal flow with two return paths. The patient phone sends to the clinic own number on the Meta Cloud API. Meta posts the message to one address at ITIG Tech, where its signature is checked, the text is encrypted and stored, and Meta is answered at once. A separate worker then writes the reply. Inside twenty four hours of the patient last message the reply can be free text. Outside that window only a template Meta has approved may be sent. The patient’sphoneThe Meta Cloud API,on the clinic’s own numberThe signature is checked,the text is encryptedand storedMeta is answered at onceA separate workerwrites the replyInside 24 hours:free textOutside it: only a templateMeta has approved
Every clinic connects its own WhatsApp Business number. There is no shared number and no SMS anywhere in this path.
How a WhatsApp message reaches the clinic A horizontal flow with two return paths. The patient phone sends to the clinic own number on the Meta Cloud API. Meta posts the message to one address at ITIG Tech, where its signature is checked, the text is encrypted and stored, and Meta is answered at once. A separate worker then writes the reply. Inside twenty four hours of the patient last message the reply can be free text. Outside that window only a template Meta has approved may be sent. The patient’s phoneThe Meta Cloud API, onthe clinic’s own numberThe signature is checked, thetext is encrypted and storedMeta is answered at onceA separate workerwrites the replyInside 24 hours: free textOutside it: only atemplate Meta has approved
Every clinic connects its own WhatsApp Business number. There is no shared number and no SMS anywhere in this path.

Limits that are built in

Rate and cost caps
At most 20 automated replies per conversation per hour and a daily model budget per clinic. When a limit is reached the conversation goes to a person rather than going quiet.
The 24-hour window
WhatsApp allows free-form replies only within 24 hours of the patient’s last message. Outside it the assistant hands over, and staff can send one approved template.
Opt-out honoured
STOP in Arabic, French, English, or Arabizi unsubscribes the patient immediately, with a written confirmation. An ambiguous word such as “cancel” gets a clarifying question instead.
Voice notes and images
Are acknowledged and passed to a person. The assistant does not listen to voice notes or open images.
Memory
Conversation text is encrypted at rest and deleted after 30 days.
Where the text goes
Replies are drafted by Anthropic’s models in the United States under commercial terms with a data processing agreement. Patient messages are not used to train models and are deleted by the provider within 30 days.

Price

As an add-on to ITIG Clinic, monthly
SoloDuoClinicPolyclinicNetworkEnterprise
+$25+$29+$39+$59IncludedIncluded
On its own, without the platform, monthly
1 doctor2 doctors3 to 5 doctors6+ doctors
$49$59$79$109

Prices are monthly, in fresh US dollars. Included on the Network and Enterprise tiers. The full pricing tables.

Talk to us on WhatsApp

Tell us how many doctors work in your clinic and which language your front desk uses. We reply in Arabic, French, or English.

WhatsApp +961 79 060 517 Open the live demo

support@itigtech.com