
Talk before you walk in. MedBrief is a voice-first AI intake agent for patients. Before a doctor's appointment, the patient has a natural spoken conversation with MedBrief, describing their symptoms, current medications, allergies, and relevant medical history. MedBrief listens, asks intelligent follow-up questions (just like a nurse would), and generates a clean, structured PDF summary the patient can hand directly to their doctor. Built for the AssemblyAI Voice Agent Hackathon, using AssemblyAI's Voice Agent API as the core conversational engine. The Problem The average primary care visit is just 15.7 minutes. Doctors interrupt patients after just 11 seconds, on average, before they finish explaining why they came in. Even when given the chance to speak, doctors interrupt patients 67% of the time. There's simply no time in a short visit to gather a full medical history, so it gets rushed, incomplete, or repeated at every visit. The Solution MedBrief moves the history-gathering conversation outside the time-constrained visit entirely. Patients complete their intake at their own pace, before they arrive, so the doctor starts the appointment already informed. How It Works: Patient starts a call in the browser, no app download, just a mic and a browser tab. AssemblyAI's Voice Agent API handles real-time speech-to-text, natural turn-taking, and spoken responses, all through a single WebSocket connection. As the patient talks, the agent uses structured tool calling to log symptoms, medications, allergies, and history in real time, with no waiting until the end of the conversation. Once the call ends, MedBrief generates a polished, clinically formatted PDF with symptoms, medications, allergies (visually flagged for safety), and history, organized into clear sections. The patient downloads the PDF and brings it to their appointment.
30 Sep 2026