To address the growing data needs, data centers are being established in new and unexplored terrains. This necessitates the development of robust data center management platforms capable of self-management. To make these systems adaptable to new uncertainties, especially in space, underwater, and remote data centers aimed at long-running or mission-critical operations, the architectures of these robust data center management platforms need to be accurately modeled to be dynamic and evolve during execution. We argue that, in addition to making systems self-managing and autonomous, there is a need for a human-centric adaptive architecture in the data center management domain. This involves incorporating human-in-the-loop mechanisms to oversee the system’s self-adaptation and control anomalies and other drifting pointers that could affect the system’s quality attributes. This paper also introduces AI Agents in the self-managing system, which can further enhance autonomous management flexibility by enabling dynamic system architecture changes through these agentic “digital workers.“ We propose a dynamic architecture framework combining AI Agents and Human-in-the-Loop for continuously improving data center management. The architecture expands the MAPE-K loop by introducing human-in-the-loop at the Plan phase and AI Agents at the Execute phase to enhance the dynamic architecture setup for autonomous self-managing data centers. This paper details the foundational AI Agents that need to be set at the Execute phase and provides a use case simulation evaluating one of the AI Agents from the proposed architecture. We also define the quality requirements, which may change at runtime in such dynamic architectures, and outline the scope of future work in this domain.