The rapid way in which e-learning platforms are evolving and materialising is testament to their ever-growing demand and their educational relevance in today's technological world. Online service providers continuously seek to optimise their products and services in their commercial endeavour to achieve a competitive edge over other e-learning providers. One way to raise the quality of service provided to online learners is through a personalised portal that provides customised academic content as well as tailored and differentiated guidance to each individual and unique student. We propose the use of companion pedagogical agents as a convenient and effective way of how to achieve the required personalised e-learning assistance. In this paper we present a case study about our research involving post-graduate students interacting with an online learning portal that recorded every student interaction. These interactions have been used as learning analytics by the underlying system to monitor student progress, propose and recommend adjustments, schedule tasks that need attention, notify the student about updates from peers, admin, and tutors, as well as ensure that the learner is maintaining a good level of participation and progress. All this information extracted from the learning analytics is not only looped back to the learners, but also to the educator in an attempt to improve the effectiveness and delivery of the entire educational process. We report on several issues encountered together with accuracy and reliability concerns as we draw a number of conclusions and offer recommendations.