Methods, apparatuses, articles, and systems for performing an installation by a client system at a time when the client system is predicted to be below a level, are disclosed. The installation may be a software or a patch. In various embodiments, the methods, apparatus et al may include performance of the adaptive prediction, and the adaptive prediction may be performed by a learning algorithm. In other embodiments, the methods et al may also develop a model of the client system's utilization by observing and recording metrics of hardware and software utilization over time.