Abstract
Artificial intelligence (AI) is playing an increasingly prominent role in medicine, and nephrology is no exception. Yet, behind this generic term lie very different realities depending on the type of data being processed. This didactic article offers a structured account of how AI works across three main data families, illustrated with concrete examples drawn from nephrology practice. With tabular data (the kind found in everyday medical records), predictive models can already anticipate acute kidney injury, intradialytic hypotension, or graft loss. With histological images, neural networks learn to detect and quantify glomerular lesions with remarkable precision, without replacing the pathologist. With text, large language models excel at reformulation, summarization, and triage tasks, more so than at diagnostic reasoning in ambiguous settings. The common thread across all three domains is the same: AI learns statistical regularities from data. Understanding this is the prerequisite for informed use: neither reflexive distrust nor uncritical delegation.