AI agents for Arabic healthcare service desks.
A hospital service desk can answer almost anything a caller asks. The one thing it must never do is give medical advice, so that refusal is configured before anything else — and ChatSheba handles departments, doctor availability, appointment times and report collection around it.
The problem in healthcare operations
Patients need departments, doctor availability, appointment times, and report and payment information without waiting on a receptionist. Reception phone load is heavy and repetitive, and a service-desk agent must never cross into medical advice.
What the AI handles
- Lists departments and on-duty doctors and books appointments by specialty and time.
- Handles report pickup, insurance acceptance, and payment questions.
- Searches a directory of facilities by specialty, location, and working hours.
- Accepts voice notes and images — prescription photos, ID cards — as input.
- Refuses diagnosis, prescription, and dosage advice and routes symptom questions to a qualified doctor.
How Sheba fits
ChatSheba runs the patient-facing service desk across WhatsApp Business API and the web, with multimodal input; custom engineering connects it to scheduling and back-office systems. Both run inside your data boundary.
Compliance and data boundaries
Each healthcare deployment defines data boundaries, escalation triggers, and permitted and non-permitted answer types. Medical-safety guardrails are part of the agent configuration, and human escalation is treated as a control, not a fallback.
- PDPL-aware by design
- On-prem and regulated-residency paths
- Per-deployment data boundaries
- Human escalation as a control
Deployment pattern
A teaching hospital and a city medical directory run Arabic AI service-desk agents. The hospital agent lists departments and on-duty doctors, books appointments, and clarifies insurance and report pickup; the directory agent searches facilities by specialty and working hours. Both refuse medical advice and route symptom questions to a doctor. Non-clinical reception load drops, and a patient asking at night about a report or an insurance card is not told to call back in the morning.
- Channels
- WhatsApp and a web chat widget
- Systems it reads and writes
- the scheduling system and back-office records
- Escalates to
- a qualified doctor on clinical queries
Part of 30+ production deployments across 14+ industries.
FAQ