Summary Digitized herbarium specimens are increasingly used to train artificial intelligence (AI) models in plant identification and other botanical applications. The abundant specimen images available in public repositories are especially amenable to AI. For instance, digitized herbarium sheets are relatively standardized – generally flattened portions of plant specimens mounted on paper with written metadata, imaged at a similar scale with uniform color‐corrected illumination. Herbarium specimen identifications rely on standardized taxonomies that have also been reviewed by one or more professionals, providing high label accuracy – a critical advantage for AI model training. In this review, we tackle the basics of AI computer vision as it relates to digitized plant specimens: how AI is applied, what hypotheses can be tested, how datasets should be constructed, and how to produce a general workflow. Lastly, we provide recommendations for best practices along with recommendations for ways that future AI researchers may refine herbarium‐focused models. In an era of declining taxonomic and specimen‐based botanical expertise, we believe that this form of AI‐based plant research presents an opportunity to augment human capacity and provides opportunity for hypothesis‐based research that must be capitalized upon.