JMIR Publications

ABSTRACT

Ambient artificial intelligence scribes and clinical decision support systems (CDSS) are usually implemented as separate digital health tools: one reduces documentation burden, while the other evaluates structured clinical data against rules or guidelines. This implementation report describes Telepatia AI Scribe and Telepatia CDSS, an integrated ambient documentation and decision support system deployed in primary care settings in Colombia and Brazil. The system captures physician-patient conversations, generates structured Spanish- and Portuguese-language medical records, maps extracted clinical content into machine-readable fields, and evaluates those fields against configured guideline-based, national, and institution-specific knowledge bases during physician note review. In a one-year implementation snapshot from April 28, 2025 to April 28, 2026, the system was deployed across 52 institutions, used by 3,123 physicians, processed 483,647 consultations, and generated 192,312 clinical decision support alerts. The architecture uses a five-layer workflow: ambient audio capture and transcription, structured medical record generation, clinical data mapping, CDSS evaluation, and monitoring with physician feedback. Its central design premise is that the scribe-generated structured note can serve as the input substrate for CDSS in environments where conventional EHR integration and structured data availability are limited. The implementation highlights practical requirements for Latin American primary care, including multilingual clinical prompting, local medication and coding mappings, heterogeneous integration patterns, clinician-facing soft alerts, and governance loops for knowledge base and model oversight. This report does not evaluate clinical effectiveness, patient outcomes, or physician behavior change. Instead, it describes the architecture, implementation choices, and operational lessons from combining ambient documentation and CDSS at scale in Latin American primary care.