Artificial intelligence (AI) enables the generation of domain-specific dialogues that can serve as teaching resources in language education. This study focuses on the Greek banking domain, a communicative context that requires learners to navigate formal registers, polite requests, and specialized financial vocabulary. Given Greek’s rich inflectional morphology, such dialogues also demand precise control of case, gender, number, and verb inflection to achieve grammatical and pragmatic accuracy. The paper outlines the rationale for selecting banking interactions and demonstrates how large language models (LLMs) can be prompted to produce dialogues for typical scenarios such as opening an account, applying for a loan, or resolving service issues. The analysis compares the output of different LLMs with respect to morphological and syntactic accuracy, register appropriateness, pragmatic naturalness, and domain-specific terminology. Results reveal variation across models: some generate fluent but overly generic exchanges, while others handle technical vocabulary well but show inconsistencies in morphological agreement and politeness conventions. Pedagogically, the study proposes morphology-aware dialogue materials for classroom use, supporting role-plays, error-spotting tasks, and targeted linguistic reflection. The findings highlight the potential of AI-generated dialogues to enrich language learning through meaningful, context-sensitive communication.
Key words: Large Language Models, data evaluation, pedagogical application, dialogue systems, banking domain