Intelligent User Interfaces (IUI) are being used in various fields of knowledge such as health, education, etc. However, developing IUI is not a trivial task in Software Engineering. Developing IUIs involves problems such as manual intervention, limited model reuse, and restricted runtime adaptability. In that context, the current research proposes to study how an artifact, designed on the basis of Model-Driven Engineering (MDE) and Machine Learning (ML), addresses the identified problems. To do that, the Design Science methodology is applied, which structures the process into three phases: problem investigation, treatment design, and treatment validation. The first phase includes a systematic literature mapping, complemented by a focused systematic review to identify the state of the art and substantiate the scientific gap. The second phase, focused on treatment design, specifies the artifact based on the knowledge gained during problem investigation. This step defines how metamodels, ML components, and deployment mechanisms interact to support the automated development of IUI. Finally, the validation phase provides empirical knowledge based on developer perceptions and a comparative analysis between the proposed artifact and traditional development approaches that rely on manual coding. The research results are expected to contribute to a technical and scientific framework that reduces manual intervention, improves development efficiency, and enables the generation of ML knowledge as input for model-driven design of IUI.