Healthcare AI Chatbot
A private hospital network operating across three facilities needed to reduce emergency department wait times and improve patient triage accuracy. We built an AI-powered triage chatbot that collects patient symptoms, assesses urgency, and routes patients to the appropriate care level before they arrive at the facility.
The Challenge
The hospital network was experiencing growing patient volumes without proportional increases in triage nursing staff. Average wait times in the emergency department exceeded 45 minutes, and the manual triage process led to inconsistent priority assignments. Patients with urgent conditions sometimes waited too long, while non-urgent cases occupied resources needed for critical care.
- Average emergency department wait time of 47 minutes before initial triage assessment
- Triage accuracy varied significantly between shifts and individual nurses, leading to inconsistent patient prioritization
- No digital pre-registration process, requiring all data collection to happen on-site
- Patient satisfaction scores for the emergency department were below the network's target threshold
- The system needed to support both Arabic and English patient interactions
- Strict healthcare data privacy requirements under regional health authority regulations
- Integration required with the existing Electronic Health Record (EHR) system for patient history lookup
Our Solution
We developed a conversational AI system that patients interact with via a web application on their mobile devices. When a patient decides to visit the emergency department, they access the chatbot through a link sent via SMS or by scanning a QR code at the facility entrance. The chatbot conducts a structured symptom assessment through natural conversation, available in both Arabic and English.
The AI engine uses a combination of a fine-tuned large language model for natural conversation and a rule-based clinical decision support system for triage classification. The LLM handles the conversational interface, understanding patient descriptions of symptoms in everyday language and asking clarifying questions. The clinical engine maps symptoms to standardized triage protocols (Emergency Severity Index) to produce a priority classification.
We built a clinician-facing dashboard in React that displays the AI triage assessment alongside the patient's conversation transcript and relevant medical history pulled from the EHR integration. Triage nurses use this dashboard to review, validate, and if necessary override the AI assessment. Every AI recommendation includes an explanation of the contributing symptoms and risk factors.
The system was deployed in Docker containers on the hospital's private cloud infrastructure to satisfy data residency requirements. All patient data is encrypted at rest and in transit, and the system maintains a complete audit trail of every interaction and clinical decision for regulatory compliance.
We conducted a phased rollout starting with one facility, using the first four weeks as a shadow mode where the AI ran alongside normal triage without affecting patient flow. This allowed us to validate accuracy against nurse assessments and refine the model before going live.
Results & Impact
- 60% — Faster Triage Time
- 94% — Triage Accuracy vs Clinician
- 32% — Reduction in Wait Times
- 4.6/5 — Patient Satisfaction Score
The AI triage system reduced the average time from patient arrival to priority assignment from 47 minutes to 18 minutes. For high-acuity cases, the improvement was even more dramatic: critical patients were identified and fast-tracked within 3 minutes of starting the chatbot conversation, compared to the previous average of 25 minutes waiting for initial nurse assessment.
After three months of operation across all three facilities, the AI's triage classifications matched clinician assessments 94% of the time. In the remaining 6% of cases, the AI tended to over-triage (assigning higher urgency than warranted) rather than under-triage, which is the clinically preferred error direction. Triage nurses reported that the pre-collected symptom data and AI recommendation allowed them to make faster, more informed decisions.
Patient satisfaction scores for the emergency department increased from 3.8 to 4.6 out of 5. Patients reported feeling more informed about their wait time and priority status. The hospital network is now expanding the system to include appointment scheduling for non-urgent cases and post-visit follow-up conversations.
Tech Stack
“The AI triage system has fundamentally changed how our emergency departments operate. Our nurses now spend their time on clinical assessment rather than administrative data collection, and our patients receive faster, more consistent care.”