AI Care Navigation for a Healthtech Startup
Intake triage and care-routing assistant that helped a healthtech startup cut time-to-appointment without replacing clinicians.
Key Details
| Challenge | Manual intake forms and chat dumped every case on nurses; wait times and no-shows climbed as the waitlist grew. |
|---|---|
| Solution | An LLM + rules care navigator that collects intake, suggests acuity and routes to the right specialty — always with clinician override. |
| Technologies | Azure OpenAI, Python, FHIR APIs, PostgreSQL |
Technologies used
Azure
OpenAI
Python
PostgreSQL
Docker
React
Client background
A Series B healthtech startup offering virtual specialty access was drowning in unstructured intake chats. Nurses spent hours clarifying symptoms; patients waited days for the right queue, and founders needed a safer AI layer that would pass partner hospital review.
Key challenges
- Free-text intake created incomplete charts and repeated nurse follow-ups.
- Routing to the wrong specialty wasted scarce clinician capacity.
- Partner hospitals required audit logs and human-in-the-loop controls.
- No-shows rose when patients waited too long without clear next steps.
What we built
- Guided intake assistant that structures symptoms, history and urgency signals.
- Routing suggestions mapped to specialty queues with clinician override.
- FHIR-oriented handoff packets into the startup’s care platform.
- Full decision log for accepted, edited and rejected AI suggestions.
Project team: 7 engineers across AI/ML, backend and product engineering — delivery over 16 weeks for a startup team that needed to ship, not slide decks.
How we delivered
Shadow
Observed nurse intake and documented refusal modes for automation.
Pilot
Launched in one specialty with clear escalation criteria.
Integrate
Connected scheduling and chart APIs with privacy controls.
Govern
Added eval sets for triage safety before expanding specialties.