The adoption of machine learning in business software is slowed down by a shortage of data science talent and challenges around efficient operationalization of machine learning models. We present IntegratedML, an embedded database capability for machine learning. This paper describes how IntegratedML provides developers with access to state-of-the-art machine learning platforms using intuitive SQL syntax. Its embedded feature extraction and algorithm selection enable fully automated model building, while model inferencing is exposed through a simple scalar function. The novelty of IntegratedML is in the deep integration into the embedding relational engine, which hides pipeline complexity from the user and guarantees high efficiencies, both at train and inference time.