Traditional assignment types, such as programming tasks focused only on correct outputs, are increasingly challenged by students’ ability to generate code using AI with minimal conceptual effort. Process-oriented assignments that highlight reasoning, design choices, justifications, and reflections provide a valuable alternative. Despite rising interest in such tasks, almost all previous research comes from high-resource environments. Little is known about implementing GenAI-mediated, process-oriented homework in low-resource settings. This paper reports an early experience with an instructor-designed AI-bot based on process-oriented assessment principles. The bot was used to record student mock interviews and interactions in a Data Structures course as part of an assignment at an African university. We found that creating this system was relatively simple, inexpensive, and capable of handling various accents at an acceptable level. This work prompts several research questions, including how much learning occurs during this type of assignment, best practices for developing AI-bots for process-oriented homework, and how to set expectations for effective learning with instructor-designed AI-bots.