The digital transformation in education presents numerous opportunities but also specific challenges, such as the growing skills shortage, the democratization of knowledge, and the increasing demand for personalized learning paths. The use of Artificial Intelligence holds promising potential to make learning processes more effective and sustainable. However, widespread integration of AI is lacking due to uncertainties about its effective application in teaching and learning. Recent research has focused on Conversational AI, which utilizes Natural Language Processing to enable human-machine interactions and is often deployed as Conversational Agents. These systems can evolve from task-oriented chatbots to companionship, offering competency-oriented support to learners. While Bloom’s Taxonomy is widely recognized for structuring traditional learning, there is a notable lack of research on how Conversational AI technology can be effectively and sustainably integrated across different competency levels. We identified 56 design features for conversational AI in education through a systematic literature review of 3891 scientific papers. We then mapped these DF to 9 overarching design principles and classified them along Bloom’s revised taxonomy. Instantiated mockups were then evaluated within the design research paradigm, contributing to the discourse on effective conversational AI in education.