The rapid advancement of artificial intelligence (AI) has introduced transformative changes across various sectors, including energy management. This paper evaluates the trust and efficiency of decisions made by stakeholders when accepting or rejecting AI responses within an AI-empowered energy software, using the VIRTSI Model—a comprehensive framework designed to assess the Variability and Impact of Reciprocal Trust States towards Intelligent systems. This evaluation is performed within an Agile development process where software products are assessed to inform the requirements of subsequent sprints. By integrating qualitative and quantitative research methods, we collected data from 19 stakeholders, including energy managers, software developers, and end-users, through 190 interactions involving preset questions. Additionally, perceived trust state dynamics automata were constructed alongside confusion matrices concerning user interactions. Our findings indicate a high level of trust in the AI-empowered system, driven by its predictive accuracy and user-friendly interface. However, challenges such as inefficiency due to perceived harmful trust states, ranging from overtrust to distrust, were identified. These evaluation results reveal significant insights into the barriers and enablers of AI adoption in the energy sector, offering practical recommendations for enhancing stakeholder trust and optimizing decision-making processes. As such, this paper contributes to a broader understanding of AI implementation in energy management and provides a foundation for future research and development in this field.