Artificial Intelligence (AI) planning systems have seen multiple challenges, in particular with regards to their architecture. Monolithic architectures lack reusability, composability, and ease of use, all of which are elements necessary to create more holistic planning systems. As the development towards general AI continues, such a holistic approach, integrating capabilities from different systems will become necessary. A need arises to decompose today’s monolithic AI planning systems into autonomous components, and integrate them in a composed planning system. We propose a systematic framework, the Monoliths to Composed AI Planning Systems Framework, which describes a process to perform this decomposition and composition of monolithic systems. The design of the framework follows considerations based in Service-Oriented Computing, and aims to build a new system founded in this paradigm. After an operationalisation step, in which we consider and choose tools, analysis approaches, and formal notations, we then examine the framework in the form of a case study. After execution of the case study we have the prototype of a composed AI planning system. The foundation of the system are the three planning systems Fast Downward, Scorpion, and DecStar-2023. The system is build using the service-oriented architecture pattern.