Manually modeling complex systems is a daunting task. Although numerous methods have been proposed for mitigating this issue, this difficult problem persists. Recent breakthroughs in generative AI and large language models have led to the creation of general-purpose chatbots, which can assist software engineers and modelers in various tasks. Still, these chatbots are often inaccurate or incorrect, and so using them in an unstructured manner might result in erroneous system models. Here, we outline a method designed for integrating chatbots into the modeling process, in a safer and more structured way. To facilitate this integration, we advocate the use of the scenario-based modeling paradigm, which has been shown to facilitate the automated analysis of models. We suggest that through the iterative invocation of a chatbot, combined with manual and automatic inspection of the models it produces, one can obtain a more robust and accurate system model. We report on favorable preliminary results, which showcase the potential of this approach.