Can AI Agents Interface Clinicians and Complex Models?
Authors: Giulio Colombini, Matteo Falcioni, Mario Massimo, Francesco Durazzi, Marianna Santoro, Paolo Tubertini, Mirko Degli Esposti, Armando Bazzani
Organizations: Dipartimento di Fisica e Astronomia “Augusto Righi”, Alma Mater Studiorum - Università di Bologna, Via Irnerio 46, 40126 Bologna, Italy. · Sezione di Bologna, Istituto Nazionale di Fisica Nucleare, Viale Carlo Berti-Pichat 6/2, 40127 Bologna, Italy. · Centre for Advanced Studies on Tourism, Alma Mater Studiorum - Università di Bologna, Via Angherà 22, 47900 Rimini, Italy. · IRCCS Azienda Ospedaliero-Universitaria di Bologna - Policlinico Sant’Orsola-Malpighi Via Giuseppe Massarenti 9, 40138 Bologna, Italy.
Background: Contemporary healthcare institutions generate daily very large amounts of data, which contain highly useful information for the improvement of work organization, planning and overall efficiency. While models and techniques to extract these insights are well known and constantly developed in mathematical biology, complex systems physics and data science, the adoption of these techniques among Clinicians and Hospital Administrators remains limited, mainly due to language and methodological divides. Objective: In this article we implement an AI Multi Agent System (MAS) that allows Users to use predictive models and make statistical analyses via natural language conversations. Methods: We developed EpidemIA, a training-free MAS endowed with a modular suite of epidemiological models written by human experts, executable on a single-GPU workstation. To avoid relaying sensitive data to third parties, the system uses locally an open-weights LLM as backbone for its two subagents. Results: Using a dataset of Emergency Department visits provided to us by IRCCS Sant'Orsola (Bologna) we test the model on a few typical use cases, demonstrating the possibility of performing both simpler statistical analyses and visualizations as well as more complicated epidemiological predictions based on dynamical models, through simple queries made in natural language. Conclusions: While still being a proof-of-concept implementation, EpidemIA showcases the transformative capabilities of LLM based technologies as an enabling factor in the widespread adoption of complex dynamical models, which is at present hindered by their steep learning curve.
Figures & tables
Figure 1: An example of AI agent architecture: an LLM backbone is given the ability to execute external functions - its tools - in a loop, to perform complex tasks.
Figure 2: Architecture of our system. The user queries EpidemIA, which accesses clinical data and epidemiological models as needed - through the analyst agent - in order to produce a clear, readable output.
Figure 3: Snapshot of the Artificial Analysis Intelligence Index as of September 13th, 2026. The LLM we employed - Qwen3.8 27B - reaches a significative level of intelligence, comparable to that of frontier open source models, at a fraction of their size.
Figure 4: Plots produced by the Agent in response to the respective User question, indicated in the caption.
Oak Ridge National Laboratory · Computational Science and Engineering Division, Oak Ridge National Laboratory, Oak Ridge, TN 37830, USA · Geospatial Science and Human Security Division, Oak Ridge National Laboratory, Oak Ridge, TN 37830, USA +1