An AI that triages patients before they reach the front desk.
Built a conversational triage assistant that collects patient history, predicts acuity level, and routes patients before they walk through the door — in 6 weeks.
The Challenge
Waiting rooms full of paperwork, not care.
ClearPath operates 22 urgent care clinics across the Southeast. Their front desk staff spent the first 15 minutes of every patient visit collecting the same information: chief complaint, medical history, current medications, allergies. All of it was re-entered into the EHR by hand.
More critically, acuity assessment — figuring out how serious a patient's condition was — happened at the front desk. Staff were not clinicians. Severe cases were sometimes routed to standard queues. Minor cases sometimes triggered unnecessary escalations.
The chief medical officer wanted to fix acuity first. We had 6 weeks.
Our Approach
A conversation, not a form.
We designed the triage flow as a structured conversation — not a web form. Patients talk to the AI the way they'd talk to a nurse. It adapts based on responses, flags warning signs, and produces a clinical summary before the patient arrives.
Clinical Workflow Mapping
Two days shadowing triage nurses across three clinics. We mapped every question they asked, the branches they took based on answers, and the signals that caused them to escalate. This became the conversation architecture.
Conversation Design & Prompt Engineering
Built the triage flow as a structured LLM conversation with guardrails. The AI collects chief complaint, symptom duration, severity, relevant history, and red-flag symptoms. It never gives a diagnosis — only a structured clinical summary and an acuity flag for the clinical team.
Prototype Build
SMS-first prototype — no app download required. Patients receive a text when they check in online, complete the triage conversation on their phone, and arrive with a clinical summary already in the queue.
Clinical Review & Safety Testing
300 synthetic patient scenarios validated against ClearPath's triage nurses. Edge cases — chest pain, pediatric symptoms, altered mental status — were hardcoded to escalate to immediate clinical review regardless of AI confidence.
Results
What we delivered.
14 min
average time saved per patient at check-in
89%
accuracy on acuity classification vs. nurse assessment
6 weeks
from first call to live prototype in 3 clinics
"The prototype went live in week 6. By week 10 we'd deployed to all 22 clinics. Our nurses spend less time on paperwork and more time on patients."
Dr. Priya Nair
Chief Medical Officer, ClearPath Urgent Care
Ready to Build?
Let's build your intelligent product.
Book a 30-minute strategy call. We'll listen to your idea, challenge your assumptions, and tell you exactly how we'd approach building it — no sales pitch, no commitment.