An Agentic AI Architecture for General Practitioners in Primary Care

Source abstract

Primary care physicians face increasing challenges in managing multimorbidity, continuous home-monitoring data, and fragmented access to specialty input. We propose an agentic artificial intelligence (AI) architecture for Medici di Medicina Generale (MMG) that combines a central planning large language model (LLM) with a bank of small, domain-specialized language models (SLMs) and determinis- tic tools. The system is designed for patient–home integration, safety triage, and transparent interoperability with regional infrastructures such as FSE 2.0 and ePre- scription. Here, we describe the functional schema and workflow of this architecture.