This paper presents a multi-dimensional taxonomy of levels of automation and reparation specifically adapted to Virtual Assistants (VAs) in the context of Human-Human-Interaction (HHI). Building from this framework, the main output of this study provides a method of calculation which helps to generate a trust rating by which this score can be used to optimise users' engagement. The authors believe that this framework could play a critical role in optimising energy efficiency in both management and consumption, particular attention has been given to the relevance of contextual events and dynamism in enhancing trust. For instance by understanding that trust formation is a dynamic process that starts before the user's first contact with the system, and continues long thereafter. Furthermore, following the evolving nature of the system, factors affecting trust and the system itself change during user interactions over time; thus, systems need to be able to adapt and evolve. Present work is being dedicated to further understanding of how contexts and its derivative unintended consequences affect trust in highly automated VAs in the area of energy consumption.