Recon
Recon is a conversation-to-chart reconciliation agent for ambient clinical AI.
Core trick
Compares transcript, generated note, and FHIR chart to flag discrepancies as clarification prompts without editing the record.
Tags
Description
Recon is a conversation-to-chart reconciliation agent for ambient clinical AI. It compares the patient conversation, AI-generated clinical note, and structured FHIR chart to identify clinically meaningful discrepancies before the note is finalized. Rather than editing the medical record, Recon provides evidence-backed clarification prompts, allowing clinicians to quickly verify potential inconsistencies while remaining in full control of the final documentation.
README excerpt
# Recon ### Conversation-to-chart reconciliation for ambient clinical AI Ambient scribes generate clinical documentation from the patient–clinician conversation. **Recon** checks whether the transcript, generated note, after-visit summary, and structured FHIR record still tell the same clinical story. It identifies a small set of potentially actionable discrepancies and proposes clarification for clinician review. It does not diagnose, modify the chart, or resolve conflicts autonomously. > **Ambient AI writes the note. Recon checks what may need clarification before the encounter is closed.** Built for **The Future of Agentic AI in Healthcare Hackathon** using Abridge's synthetic ambient-FHIR dataset. ## Example **Patient says** > "I stopped taking lisinopril about six months ago." **FHIR medication list** > Lisinopril 10 mg — active **Generated note** > Current medications include lisinopril. **Recon surfaces** > **Possible stale medication record** > The patient reports discontinuing lisinopril, but it remains active in the structured medication list and generated note. > > **Suggested clarification:** > "Can we confirm whether you are currently taking lisinopril?" The clinician decides whether any update is appropriate. ## Why Recon Ambient documentation systems are optimized to capture and summarize the current encounter. They may not independently reconcile every statement against the patient's longitudinal record or verify consistency across downstream