The problem addressed in this paper is to define how human skills can be improved to match the new trends in Industry 5.0. Several technologies can help us to develop enough learning to solve complex tasks, such as the virtual approaches to iterative simulation in immersive environments. The Digital Twin (DT) concept has appeared more frequently in the industrial scenario. Thus, the training laboratories must follow technological development and be more attractive to engineering students. From this scenario, the objective is to create shared experiments at a low cost by proposing a framework to be adapted to training in the context of a digital twin operation. As a result, we exemplify the main concepts of operating a didactic robotic arm for remote access to both the real side and the corresponding digital twin. We simulated virtual AI-based students learning an activity and evaluated the learning curve of an experimental example. The contribution of this research deals with a virtual mirrored model of a real process as an important strategy for training because the user can try many times and explore alternative ways to solve a problem and learn from errors by themselves without interfering with the real side of a process.