Cultivating AI leaders: the role of clinical informatics fellowship in transforming healthcare

On behalf of the American Medical Informatics Association (AMIA) Community of Clinical Informatics Fellowship Program Directors, representing over 60 accredited academic programs across the United States, we commend Mahajan and Bates for articulating the urgent need for physician leaders trained to steward artificial intelligence (AI) responsibly in healthcare.¹ We fully concur that clinical oversight of AI systems is essential for patient safety and ethical implementation.

We further emphasize that Clinical Informatics (CI) Fellowship training² already provides a comprehensive and well-established pathway for preparing physicians for this evolving responsibility. Since its recognition as an ACGME-accredited subspecialty in 2014, CI training programs have demonstrated remarkable agility—adapting curricula to meet the demands of emerging AI technologies, including machine learning (ML), deep learning, and large language models (LLMs). Just as radiology or pathology matured through technological transformation, CI has continually evolved from its foundational domains—health information systems, decision support, and change management—to encompass data governance, advanced analytics, and applied AI competencies.

Across programs, fellows receive structured education in data literacy, programming fundamentals, model development and validation, risk monitoring for model drift, and the ethical use of generative models within care delivery. They gain experiential training in evaluating vendor algorithms, integrating AI tools into clinical workflows, and implementing governance frameworks such as model cards and post-deployment surveillance. A survey of CI fellows and graduates from 2016 to 2024 found that 36% even developed their own ML models during training.³ More importantly, CI fellows are trained to begin with clinical problem definition—identifying where algorithmic interventions add value—before pursuing technical solutions. The foundational CI knowledge of the healthcare domain, health information systems, and the skills of process improvement, workflow optimization, implementation and change management are just as critical, or one might argue, even more so, for AI applications.

This rapid curricular evolution ensures that graduates are not only conversant with AI theory but are capable of real-world application, oversight, and leadership within complex health systems. This is further supported by numerous articles co-authored by CI fellows and graduates in recent years, including implementation frameworks for safe and effective AI deployment, real-world evaluations of predictive models and LLMs, and applied studies on AI-enabled clinical workflows.4, 5, 6, 7, 8, 9 In this regard, Clinical Informatics Fellowships already cultivate the “Clinical AI Physician” envisioned by Mahajan and Bates—physicians equipped to critically appraise, implement, and manage AI safely and equitably.

As artificial intelligence rapidly reshapes medicine, strengthening and expanding Clinical Informatics training is a strategic imperative. Clinical Informatics programs already design and deliver core informatics curricula for medical students, residents, and other trainees, embedding these competencies across clinical education rather than creating redundant silos. The infrastructure, accreditation standards, and national community already exist to support this next generation of physician leaders.

Contributors

VL and JDH conceptualized the commentary. VL wrote the original draft and all authors reviewed and critically appraised the work.

Declaration of interests

The authors have no financial conflict of interest to declare. All authors have a volunteer (unpaid) position in the American Medical Informatics Association specifically related to supporting clinical informatics fellowships. ML is the owner of Zimi Medical Technologies LLC.