E-Learning is often the equivalent of a conventional text-book: Identical for all learners. Instead, it should be more like a classroom, where a teacher adapts some, but not all, elements of teaching for each learner individually. For this, interests of students must be identified. An online learning platform was enhanced through agents to derive keywords from the material contained and identify a sub-set of interesting ones for each learner based on his/her actions. These are then used for automatic notifications on interesting events. Key findings from the first course held with support of this system are presented in this paper and possibilities for further enhancements discussed.