The quest for the efficient and autonomous placement of network services is crucial to progressing towards a fully automated network, commonly referred to as a “zero-touch network”. In this presentation, we focus on the results of our research into exploiting the potential of deep reinforcement learning strategies in the field of network slicing. Our aim is not only to elucidate the remarkable advantages these techniques offer over conventional methods but also to address their inherent limitations. In addition, we will highlight new avenues that we are currently exploring to enhance the reliability aspects of network slicing when deploying techniques based on deep reinforcement learning. During this presentation, we will outline some of the solutions and tangible results we have obtained in the pursuit of this critical mission.