Intelligent tutoring systems collect learners’ traces in tech-nology-enhanced learning environments with the aim of guiding and improving their learning in real time. Research has succeeded in developing data models that optimize the prediction of learning outcomes. Accurate prediction, however, does not provide information on how to achieve the desired learning outcomes. Recent approaches emphasize an interdisciplinary design process using human-computer interaction and learning engineering methods. Accordingly, this paper introduces an adaptable assistance system for vocational education that is developed in an interdisciplinary collaboration between learning and computer science experts. The assistance system supports both the processes of self-regulated learning and collaborative knowledge building by enabling learners to individually choose from topic-specific and/or interaction-specific recommendations. The chatbot recommendations are derived from a learning suggestion middleware that evaluates xAPI statements. It is based on explanatory learner models that provide not only accurate predictions, but also interpretable and actionable insights into learners’ activities and their learning process. A graphical knowledge structure provides an overview of the learning content, learner’s progress and allows free navigation. Ideas on how to evaluate the assistance system in scenarios of self-regulated learning and workplace learning will be outlined.