Early phases of IT project management require well-structured plan and scope documents, which are often created manually and prone to inconsistencies. This paper presents MAPS-AI, an AI-assisted tool for model-driven generation of project initiation artefacts. It transforms high-level conceptual models into structured plans and scopes, combining large language models with model transformation rules. MAPS-AI integrates modelling, AI-assisted text generation, and prompt engineering techniques to enhance content relevance and consistency. This paper demonstrates the tool prototype, showcasing its potential to improve content traceability in early project phases.