One strategy for intelligent agents in order to reach their goals is to plan their actions in advance. This can be done by simulating how the agent’s actions affect the environment and how it evolves independently of the agent. For this simulation, a model of the environment is needed. However, the creation of this model might be labor-intensive and it might be computational complex to evaluate during simulation. That is why, we suggest to equip an intelligent agent with a learned intuition about the dynamics of its environment by utilizing the concept of intuitive physics. To demonstrate our approach, we used an agent that can freely move in a two dimensional floor plan. It has to collect moving targets while avoiding the collision with static and dynamic obstacles. In order to do so, the agent plans its actions up to a defined planning horizon. The performance of our agent, which intuitively estimates the dynamics of its surrounding objects based on artificial neural networks, is compared to an agent which has a physically exact model of the world and one that acts randomly. The evaluation shows comparatively good results for the intuition based agent considering it uses only a quarter of the computation time in comparison to the agent with a physically exact model.