Improved Intake and Triage by 35% with an AI MVP for a Saudi Health Startup

The startup’s doctors and care coordinators were spending too much time on calls gathering basic information before each consultation. Patients repeated the same symptoms and history across channels, while appointment queues stretched longer than necessary. Many potential patients dropped off during registration, frustrated by long forms and delays.

They wanted a solution that collected structured information before appointments, guided patients to the appropriate consultation type, and provided doctors with a clear summary before the call started.

• Unstructured symptom descriptions were submitted in free text, which made it hard to standardize triage.
  
• Overloaded medical staff handling both administrative intake and clinical care.
  
• Difficulty prioritizing urgent cases versus routine consults, leading to poor slot utilization.
  
• The need to respect privacy, security, and regional regulatory expectations while introducing AI into patient workflows.

• Our project team initiated an AI audit and telemedicine workflow study. Together with the medical and operations teams, the team identified the most repetitive and time-consuming steps in the intake journey and mapped them into a safe, structured AI flow.
  
• We designed and developed an AI-powered triage and intake assistant that collected symptoms, duration, basic history, and medication details through conversational prompts powered by LangChain and OpenAI models.
  
• The team converted patient responses into structured triage summaries that followed clinical guidelines defined by the client’s medical leads.
  
• Integrated with the existing telemedicine platform through secure APIs, attaching triage summaries directly to the doctor’s appointment dashboard.
  
• The backend was built using Python and FastAPI for AI services, Node.js for integration layers, and PostgreSQL for structured data. All workloads were containerized with Docker and deployed in a secure cloud environment that met the client’s data residency requirements.

The AI-powered MVP significantly reduced patient friction and freed up valuable doctor and nurse time. Our solution safely expanded in subsequent phases.

• 93% completion rate for AI-guided triage sessions among patients who started the flow.
• 3x faster patient intake compared to the previous manual process, which reduced waiting times for virtual consultations.
• 47% reduction in manual triage time for clinical staff, enabling them to focus on complex, high-risk cases.
• 38% improvement in doctor slot utilization, achieved by scheduling patients in a more structured, prioritized manner.