Scribe Sign-Off Cross-Check: Automated Accuracy Flags for AI-Drafted Notes

A QA layer that cross-checks every ambient-scribe draft note (DAX Copilot, Nabla, etc.) against the patient's structured EHR data and flags clinically significant discrepancies for the physician to resolve before signing. New evidence strengthens the case — LLMs extract discrete clinical assertions from notes with high accuracy, and note-vs-chart medication mismatches run around 23% in one study — but also bounds it: structured EHR data is itself often stale, so flags mark discrepancies between two fallible sources rather than proven scribe errors, and flag volume must stay low enough not to erode the scribe's time savings.