There have been many attempts to integrate automated planning and rational agents. Most of the research focuses on adding support directly within agent programming languages, such as those based on the Belief-Desire-Intention model, rather than using off-the-shelf planners. This approach is often believed to improve the computation time, which is a common requirement in real world applications. This paper shows that even in complex scenarios, such as in the Multi-Agent Programming Contest with 50 agents and a 4 s deadline for the agents to send actions to the server, it is possible to efficiently integrate agent languages with off-the-shelf automated planners. Based on the experience with this case study, the paper discusses advantages and disadvantages of decoupling the agents from the planners.