Artificial intelligence in nephrology: recent advances

Artificial intelligence (AI) is moving rapidly from experimental prediction models toward workflow-integrated tools that can support clinical care, education, research, and operational efficiency. Kidney care is particularly well suited to AI because it is data-rich, longitudinal, and multidisciplinary, incorporating laboratory trajectories, medications, imaging, pathology, dialysis machine data, transplant immunology, wearable sensors, and unstructured clinical notes. Recent advances include electronic health record-based models for acute kidney injury prediction, machine-learning approaches for chronic kidney disease progression and diabetic kidney disease risk stratification, automated measurement of total kidney volume in autosomal dominant polycystic kidney disease, quantitative renal pathology, intradialytic event prediction, transplant allograft risk models, AI-enhanced electrocardiography for electrolyte and hypertension-related applications, and large language models for documentation, summarization, patient education, and clinical decision support. However, high discrimination in retrospective datasets has not consistently translated into improved outcomes after clinical deployment. Evidence from AKI alert trials emphasizes that AI must be coupled to actionable workflows, clinician oversight, calibration, and monitoring for bias, safety, and alert burden. Generative AI offers a compelling opportunity to reduce administrative work and restore clinician time, but hallucination, incomplete guideline concordance, and variable safety remain barriers. The next phase of AI in nephrology should prioritize pragmatic validation, transparent governance, human-AI collaboration, and practical applications that improve patient-centered kidney outcomes rather than technology adoption for its own sake.