IOPT Petri nets are commonly used as a formalism for modeling digital control systems, and the existing tools are capable of verifying, simulating, and deploying the implementation of these systems. Low-level Petri nets are sufficient for simple use cases to express the system’s behavior. However, using low-level nets quickly becomes challenging to manage for more complex cases, such as multi-agent systems. Furthermore, there is no easy formalism to express the interactions of these systems with Artificial Intelligence systems. To avoid node explosion and net complexity, characteristics of low-level approaches, and the use of high-level nets, with their capability to hold complex data in their tokens, contribute to reducing the net complexity and improving legibility. This leads to the need to develop tools for designing, validating, simulating, and deploying these high-level Petri nets across various target platforms, enabling the construction of cyber-physical multi-agent systems with AI capabilities. This paper outlines some challenges in modeling such systems, focusing on agents with decision-making processes that incorporate artificial intelligence, and proposes a high-level extension to the low-level IOPT net class, aiming for more compact models and more straightforward implementation of such systems.