AI & Communication

It reads 'انا عاوز اجى بكرا الصبح'

Trained on how Egyptian patients actually type

Egyptian dialect
انا عاوز اجى بكرا الصبح
→ AI booked morning slot
3ayez a7gz maw3ed
→ Understood franco-Arabic
is dr free monday
→ Answered availability

The problem

Off-the-shelf AI is trained on formal Arabic and stumbles on how Egyptians actually type on WhatsApp - misspelt words, franco-Arabic, slang, and half-sentences that a real receptionist would understand instantly.

Egyptian Dialect

Egyptian patients don't write in formal Arabic. They write in a mix of colloquial Arabic, franco (a7gz maw3ad), English words in the middle of Arabic sentences, and typos that come from typing fast on a phone. Off-the-shelf AI - even the good ones - fails on this constantly, and each failure means a lost booking.

Kashfy's AI is trained on tens of thousands of real Egyptian patient WhatsApp conversations. It reads 'انا عاوز اجى الاسبوع الجاي' the same way a receptionist would. It handles '3ayez a7gz' and 'عندي وجع في راسي' with equal ease. It understands 'الدكتور موجود بكرا؟' without needing the patient to phrase it politely.

When it does need to escalate - say the patient wrote something in a dialect Kashfy hasn't seen before - it hands off with the full conversation, not a chopped 'I didn't understand' message. Reception picks up where the AI got confused, resolves it, and Kashfy learns from the resolution.

Inside

  • Handles Egyptian colloquial Arabic natively - not translated MSA
  • Understands franco-Arabic ('3ayez a7gz maw3ad')
  • Reads through common typos and phone-typing shortcuts
  • Fluent in common Egyptian medical shorthand ('ضغطي عالي')
  • Switches to formal Arabic or English when the patient does
  • Learns from every human correction reception makes
  • Handles voice notes: transcribes and responds
  • Understands mixed sentences ('عندي allergy من penicillin')

Who this is for

A

Clinic in a working-class neighbourhood

Patients use heavy colloquial, franco, and voice notes. Formal-Arabic AI fails here. Kashfy is the only one that understands them.

B

Clinic serving a mix of Egyptian and expat patients

Patients write Arabic and English in the same message. The AI handles both without asking the patient to 'pick a language'.

C

High-volume WhatsApp with voice-note patients

Older patients prefer sending voice notes. Kashfy transcribes them accurately in Egyptian dialect and responds in text they can read.

Common questions

Can I test it with my own dialect?

Yes - visit /try and chat with the AI directly. Send it real messages the way your patients would, and see how it responds.

What if the AI still doesn't understand something?

It hands off to reception with the full context. Your correction becomes training data - the same phrase won't confuse it again.

Does it work with voice notes?

Yes. Voice notes are transcribed on-device using Egyptian-tuned models, and the AI responds to the transcription.

Can I make it more or less formal?

Yes. There's a tone slider in the knowledge base - casual, professional, or formal - that shapes every reply.

Ready to run egyptian dialect on Kashfy?

First 50 bookings free. No card, no clock.

·