Most health systems have AI policies in place. But as AI becomes embedded in daily workflows, a growing share of use is happening quietly — outside formal approval pathways, and often beyond leadership’s visibility.

“99% accurate” may sound decisive, but in healthcare AI it tells only a fraction of the story. Real impact is determined less by model performance alone and more by whether organizations can operationalize the output within their own workflows, teams, and constraints.

In this interview, Dr. Alexandra Rabotin shares what ambient AI scribes look like in real clinical workflows, from bilingual encounters to safety-net settings.