The healthcare landscape is currently witnessing a paradigm shift where artificial intelligence (AI) is transitioning from a high-tech novelty to a fundamental clinical necessity. This evolution was the central focus of a high-profile session at ENDO 2026, the annual meeting of the Endocrine Society, held in mid-June. Titled "Artificial Intelligence in Endocrinology: Practical Uses, Lessons Learned, and What Comes Next," the session provided a comprehensive roadmap for clinicians navigating the complexities of digital transformation. Led by prominent researchers from the Mayo Clinic and the University of Florida, the presentation underscored a critical message: while AI will not replace physicians, those who integrate these tools into their practice will inevitably surpass those who do not.
The Convergence of Technology and Clinical Practice at ENDO 2026
The ENDO 2026 session, held on Saturday, June 13th, drew a capacity crowd of endocrinologists, researchers, and healthcare administrators. Chaired by Juan P. Brito, MD, MSc, a professor of medicine and director of the Care and AI Laboratory at the Mayo Clinic, the session was designed to bridge the gap between theoretical AI potential and immediate clinical utility. Dr. Brito, who also serves as the innovation and quality chair for Mayo’s Division of Endocrinology, emphasized that the goal was to provide a balanced perspective—merging the practical tools available today with the systemic lessons learned from early implementation.
The session was co-chaired by Camilo Daniel Gonzalez, MD, PhD, of the Autonomous University of Nuevo León in Monterrey, Mexico, highlighting the international interest in standardizing AI protocols within the specialty. The dialogue centered on how AI can be leveraged to address the chronic shortage of time in modern medicine, allowing clinicians to refocus on patient-centric care rather than administrative burdens.
Everyday AI: Enhancing Efficiency and Reducing Cognitive Load
David Toro-Tobón, MD, a Mayo Clinic Scholar and assistant professor of medicine, opened the session by introducing the concept of "Everyday AI." This category refers to generative AI and large language models (LLMs) that can be integrated into the current medical workflow without requiring massive infrastructure overhauls.
Dr. Toro-Tobón categorized these tools into three primary functional areas:
- AI Scribes and Documentation Support: Tools that listen to patient encounters and generate structured clinical notes, significantly reducing the "pajama time" clinicians spend on electronic health record (EHR) entries.
- Point-of-Care Decision Support: Platforms like OpenEvidence and other clinical data retrieval tools that allow physicians to query vast databases of medical literature for real-time evidence-based answers.
- Administrative and Research Support: AI-driven assistance for drafting patient communications, summarizing long medical histories, and organizing data for academic publications.
The primary value proposition of Everyday AI is the reclamation of time. By offloading the cognitive load of routine tasks to machines, endocrinologists can dedicate more energy to the "human" aspects of medicine—understanding patient values, navigating complex psychosocial factors, and performing physical examinations. However, Dr. Toro-Tobón cautioned that the use of these tools requires a sophisticated understanding of their limitations, particularly the risk of "hallucinations," where AI generates plausible-sounding but factually incorrect information.
The Privacy Imperative: HIPAA vs. Business Compliance
A significant portion of the discussion was dedicated to the legal and ethical nuances of data privacy. Dr. Toro-Tobón highlighted a common and dangerous misconception regarding "business-compliant" AI tools. Many health systems utilize software suites that feature a "green shield" or similar branding, indicating that the data is protected for corporate use. However, business compliance does not automatically equate to HIPAA (Health Insurance Portability and Accountability Act) compliance.
Business-compliant tools ensure that the data entered is not used to train the developer’s public models, but they may not meet the rigorous security, auditing, and data-handling standards required for Protected Health Information (PHI). Dr. Toro-Tobón noted that while individual tools like Doximity’s "Ask" or OpenEvidence may offer HIPAA-compliant pathways via National Provider Identifier (NPI) verification, these arrangements are often personal. Because patient data technically belongs to the healthcare institution, clinicians must ensure their specific institutional contracts and configurations are explicitly approved for clinical use.
Implementation Science: The Three Pillars of AI Adoption
Naykky M. Singh Ospina, MD, a professor at the University of Florida, shifted the focus to the systemic challenges of AI adoption. Her research in implementation science identifies why many promising AI tools fail to gain traction in the real world. She organized these challenges into three distinct pillars:
The Technical Pillar
AI models must be mature enough to handle the specificities of medical inquiry. Unlike general-purpose AI, medical AI must account for longitudinal data, clinical uncertainty, and the varying contexts of individual patient histories. A model that performs well in a controlled laboratory setting may struggle with the "messy" data found in standard clinical practice.
The Clinical Pillar
Technological accuracy is insufficient if it does not lead to improved patient outcomes. Dr. Singh Ospina argued that the medical community must demand evidence that AI tools actually reduce mortality, improve quality of life, or increase the efficiency of care delivery. Without proven clinical utility, AI remains an expensive distraction.
The Implementation Pillar
This involves the "social" side of technology—building trust among patients and clinicians and integrating tools seamlessly into existing workflows. If an AI tool requires a physician to open a separate window or perform additional clicks, it is unlikely to be adopted, regardless of its accuracy.
Case Studies in Success and Stagnation
To illustrate these pillars, Dr. Singh Ospina compared two high-profile AI applications in endocrinology: diabetic retinopathy screening and thyroid ultrasound assessment.
Diabetic Retinopathy: A Success Story
AI-driven screening for diabetic retinopathy has successfully cleared all three hurdles. FDA-approved tools now allow for immediate retinal imaging in primary care clinics. The AI provides a binary "refer/no refer" result on the spot, overcoming common patient barriers such as transportation and scheduling conflicts with ophthalmology specialists. This tool succeeds because it is integrated into the workflow, backed by clinical guidelines, and supported by specific billing codes.
Thyroid Ultrasound: A Work in Progress
In contrast, AI tools for thyroid ultrasound assessment face significant implementation barriers. While several models can flag suspicious nodules, they have yet to prove that they improve overall patient care or reduce unnecessary biopsies. Furthermore, clinicians remain concerned about the models’ performance on rare thyroid cancer subtypes, as most AI training sets are dominated by common papillary thyroid carcinomas. The complexity of thyroid management—which involves balancing patient preferences, surgical risks, and long-term monitoring—makes it a more difficult "niche" for current narrow AI to conquer.
Supporting Data and the Economic Context
The urgency discussed at ENDO 2026 is supported by broader industry trends. According to data from the American Medical Association (AMA), nearly 63% of physicians report symptoms of burnout, with administrative burden cited as a leading cause. AI scribes have shown the potential to reduce time spent on EHR documentation by up to 50%, a statistic that aligns with Dr. Toro-Tobón’s emphasis on "giving time back" to clinicians.
Furthermore, the global market for AI in healthcare is projected to reach over $180 billion by 2030, with a significant portion of that investment directed toward diagnostic imaging and chronic disease management. In endocrinology specifically, the prevalence of diabetes—affecting over 500 million people globally—creates an overwhelming demand for scalable screening and management tools that human clinicians alone cannot meet.
Chronology of AI Evolution in Endocrinology
The path to the ENDO 2026 session can be traced through several key milestones:
- 2022-2023: The "LLM Boom" led by ChatGPT sparked initial curiosity and concern among medical professionals regarding generative AI.
- 2024: Professional organizations, including the Endocrine Society, began forming task forces to establish ethical guidelines for AI use in clinical research and practice.
- 2025: Leading institutions like the Mayo Clinic launched specialized Master’s programs in AI for Healthcare, aimed at creating "clinician-scientists" who possess both technical and medical fluency.
- 2026: The focus shifted from "What is AI?" to "How do we safely and ethically implement AI?" as seen in the sessions at ENDO 2026.
Broader Impact and Future Implications
The consensus among the experts at ENDO 2026 was that AI represents a "basic science" for the modern era—a new set of techniques that every clinician must eventually master. The development of multidisciplinary teams—comprising data scientists, clinicians, patients, and implementation experts—is essential from the earliest stages of tool design.
Dr. Brito concluded the session by noting that while AI holds tremendous promise, the field must maintain rigorous standards. The transition from "exploration" to "thoughtful implementation" requires a cautious approach to tools that directly influence diagnosis and treatment. In the near term, the most significant impact will likely remain in the "unglamorous" parts of medicine: routine cognitive tasks and administrative streamlining.
As the session adjourned, the overarching sentiment was one of cautious optimism. The endocrinology community is no longer asking if AI will change the field, but rather how they can steer that change to ensure it remains patient-centered, evidence-based, and ethically sound. The "AI revolution" in medicine is not a sudden disruption, but a disciplined evolution toward a more efficient and human-centric healthcare system.
