[AI-powered scribe : benefits, limitations and implementation].

Abstract

Clinical documentation accounts for an increasing proportion of clinicians' working time, reducing the time devoted to patients and affecting personal satisfaction. In this context, artificial intelligence-powered transcription systems, also referred to as ambient scribing, are emerging as potential solutions to reduce documentation burden. Several studies suggest a reduction in documentation time and a perceived improvement in clinicians' experience. However, these tools also have limitations, including risk of errors or generation of inaccurate content, necessitating systematic human validation. Their adoption should be embedded within a structured approach that considers integration with electronic health records, performance, confidentiality, data security, and regulatory framework.