Arabic AI Medical Scribe — Draft Clinical Notes You Review and Sign

Picture a dermatologist in Riyadh seeing 28–35 patients a day, consulting in Najdi Arabic, switching mid-sentence into English clinical terms — “we'll start isotretinoin, five-milligram, review the LFTs” — with the occasional patient replying in Franco-Arabic over the clinic's WhatsApp. Her clinic-management system requires typed English notes for insurance, so she either types during the consult, eyes on the keyboard instead of the patient, or stays 60–90 minutes after the last appointment reconstructing the day from memory — which is exactly when the clinical detail dies.

She tried a dictation app: it transliterated the dialect into nonsense and dropped every drug name. She tried a global AI scribe trial: impressive in English, useless the moment the patient spoke Arabic. Three failure modes, one root cause — typing during the consult breaks eye contact and thins the notes; dictation apps mangle dialect and force doctors to speak like a machine; global scribes are dead on arrival with any Arabic, English, or Franco mix. The problem was never the doctor. The tools were never built for how her consultations actually sound.

The flow is four steps, all in the room. First, consent at reception — the patient agrees before anything records, and a declining patient is documented the traditional way. Second, the scribe listens: no wake words, no dictation grammar, just natural speech in any mix of Arabic, English, and Franco. Third, the draft appears in SOAP or your clinic's own template, before the patient even reaches the front desk. Fourth, you review, edit, and sign — and that edit-time is logged automatically, becoming the minutes-saved report your pilot is measured on.

The day-30 report shows baseline minutes per consultation, pilot minutes per consultation, minutes saved per day, and the percentage of notes signed same-day — all measured, none of it invented. The principle is stated plainly, in both languages, because it is the whole trust model: the scribe drafts, you decide, and a note without your signature does not exist.

One recording can contain Gulf or Egyptian dialect, English medical terminology, and Franco-Arabic code-switching in the same sentence — and still produce a structured note in your clinic's format. Global scribes in the Abridge/Nuance tier have near-zero dialect handling; dictation apps require the doctor to change how they speak. We publish specialty golden-set evaluations built from consented, de-identified recordings, versioned and re-tested on every model change, and we launch one medical specialty at a time so the vocabulary is deep rather than shallow — depth over breadth is the product strategy, not a limitation we're hiding.

Roughly 95% of global AI vendors cannot handle dialectal Arabic, and none publish Arabic clinical evaluations. Augnito's Arabic scribe is deployed at Almoosa Health, a hospital system — enterprise vendors need hospital contract sizes and structurally skip private clinics. From SAR 300/month per doctor, this is priced for the tier nobody else serves.

MIT NANDA's research found that 95% of enterprise gen-AI pilots produce no P&L return — which is exactly why this pilot is instrumented from day zero, not measured after the fact. Week 0 (Assess) times your current documentation on five real consultations and installs the consent kit at reception. Days 1–14 (Pilot) tune the scribe to your specialty vocabulary and note format while you use it on real consultations with full review-first control. Days 15–45 (Prove) continue instrumented use and produce the day-30 minutes-saved report per doctor — the guarantee gate. After the pilot (Scale), more doctors and branches can join, with a second specialty only after retention proof. Monthly (Operate), terminology additions come from your own edits, evals re-run on every model update, and you receive a fresh minutes-saved report every month.

This is a pilot-program page, not a live storefront: cohort 1 opens in November 2026 in one medical specialty, with limited clinic seats so every pilot clinic gets real engineering attention. It is a pure subscription — no setup fee, no separate operations retainer. The pilot month is billed at your tier's price and refunded in full under the guarantee below. Joining the waitlist locks in pilot pricing and commits you to nothing.

Solo Doctor is SAR 300/month and includes one specialty, draft notes plus the review app, PDF/print export, and a monthly minutes-saved report. Clinic (2–10 doctors) is SAR 500/month per doctor and adds the clinic dashboard, custom note templates, and structured copy-to-EHR export. Chain/Multi-branch (10+ doctors) is SAR 800/month per doctor and adds EHR integration scoped per system, an SLA, and multi-branch reporting. Egyptian clinics join pilots on a separate EGP pricing band; UAE clinics are invoiced in AED equivalents.

Patients consent at reception, in Arabic and English, before any recording begins — a patient who declines is simply documented the old way, with zero pressure. Audio and notes are hosted regionally, encrypted in transit and at rest, and accessible only to the treating doctor and named clinic admins. Audio is deleted on a contractual schedule after the note is signed, and patient data is never used to train public models. Deployments are designed around Saudi PDPL, UAE data protection law, and Egypt's Law 151/2020, with a data-processing agreement included in every contract.

This product documents consultations. It does not diagnose, recommend treatment, or triage patients — by design and by contract. That boundary is what keeps the scribe a documentation assistant, never a clinical decision-maker.

Before any pilot case study exists, we show you the real artifacts: a 2-minute product walkthrough of a simulated consultation (labeled as such), a minutes-saved report template populated with clearly labeled sample data, and a description of our eval methodology. For category context — not our own results — Middle East healthcare AI is projected to grow from $436M in 2024 to $8.4B by 2033 (Grand View, vendor estimate), Augnito's Arabic scribe has been live at Almoosa Health since January 2025, and 21Doctors raised a pre-seed round in September 2025. We never invent clinic names, doctor names, or outcomes — until a pilot clinic consents in writing, every proof slot renders the anonymized template, clearly labeled.

At launch, the scribe works alongside any clinic-management system or EHR: the signed note exports as structured text matched to your template, as PDF, or for print, and most clinics paste it into their existing system in seconds — a step the minutes-saved measurement includes, so the reported number is honest end-to-end. Direct EHR integrations are scoped per system for chains through our AI Integrations engineering practice; we name no vendor partnership that doesn't exist. On the model layer, we benchmark frontier multilingual speech models against fine-tuned options on our own specialty eval set, version every prompt, and run automation on n8n where clinic workflows need it. If your front desk is also drowning in missed calls, the scribe pairs with our Arabic Voice AI Agents — one clinic stack, from the phone call to the signed note.