Industrial real-time embedded systems run complex software applications with strict timing constraints to ensure efficient operation. Increasing computational demands often strain device resources, leading to missed deadlines and disrupted task execution. This paper proposes a novel decision-making mechanism for offloading soft real-time tasks to edge or cloud servers. The mechanism contains two major components: (i) a monitoring agent to observe and collect task-level performance parameters during run-time, and (ii) a decision-making component to decide, based on the collected parameters, which tasks should be offloaded to improve the overall system performance. We implement the proposed mechanism in the FreeRTOS real-time operating system to provide evidence of its feasibility. We also perform a set of experiments to show performance of the proposed mechanism. The results demonstrate that the proposed mechanism ensures predictable execution of the system with minimal overhead, offering a scalable solution for resource-constrained soft real-time systems.