To foster their full integration into the mainstream of society, speech-impaired individuals need to be empowered to use vocal language in various daily life situations. AI-based AAC systems enable speech-impaired individuals to generate completed utterances for use in conversations. However, available solutions exhibit remarkable drawbacks. They do not meet all posed requirements and cannot provide completed sentences that match the expected styles of formal and informal conversational contexts. Therefore, design requirements for a Generative AI-based utterance composition approach were identified. It has been followed design science research to derive design principles and an instantiation of an AAC prototype that enables context-specific and user-individualizable articulation. The artifact was demonstrated and evaluated with formal and informal interactions commonly performed in the daily routine of visiting a restaurant. In particular, the importance of meeting all specific requirements and employing Generative AI, such as ChatGPT-4.0, to generate both formal and informal utterances, is exemplified.