Grown custom ABAP code has an architecture — it is just rarely written down anywhere. There are tools to define or check architectures, e.g. ABAP package interfaces and the accompanying package checks. In practice, not every team uses them, and hardly anyone retrofits them into code that has grown for decades. So nobody can say with confidence how well their custom code is structured internally or towards the standard, e.g. in terms of Clean Core.
In the session, we'll take the abapGit project as an example code base. We'll use an AI agent to derive a component architecture for it from the code and its dependencies. The first result helps us to get an overview of this code base. Then, we'll refine the model step by step, partly by hand in a graphical editor, partly by prompting the AI again. The resulting model can then be used to continuously check all custom code and warn us when we introduce undesired dependencies.