Inspiration
Clinics in under-resourced areas lose patients to something completely preventable: call overload, delayed intake, and missed follow-ups. A missed call can mean a missed diagnosis. We wanted to fix the front desk before trying to fix anything more complex.
What it does
Clinic Copilot is an AI voice agent that acts as a clinic's front desk in both directions. It answers inbound calls and handles multilingual patient intake, triages symptoms by urgency with clear safety disclaimers, escalates emergencies immediately, books appointments based on urgency and availability, and makes outbound follow-up calls after visits to check on recovery and catch missed care early. It also sends reminders to reduce no-shows.
How we built it
This is a system design and architecture submission for the Idea Round. The approach is a voice AI layer, speech-to-text, LLM reasoning, and text-to-speech, connected to a telephony API for both inbound and outbound calls, with a lightweight backend to manage triage logic, booking, and call logs.
Challenges we ran into
Balancing usefulness with safety was the hardest part. The agent needs to be genuinely helpful for intake and triage without ever crossing into giving medical advice. Getting the escalation logic right, so it reliably catches anything urgent, was the core design challenge.
Accomplishments that we're proud of
Designing a system that solves two real clinic problems at once, missed calls and missed follow-ups, in a single deployable agent instead of two separate tools. Also proud of building in safety and escalation logic from the start rather than as an afterthought.
What we learned
That the biggest barrier to healthcare AI adoption is not the technology, it is trust. A tool like this only works if patients feel heard and clinics feel confident it will never overstep into medical advice.
What's next for Clinic Copilot
Building a working prototype with a real telephony integration, piloting it with one clinic to test the triage and booking flow end to end, and refining the multilingual intake based on real patient conversations.
Log in or sign up for Devpost to join the conversation.