Abstract
This paper demonstrates using large language models (LLMs) to extract insights from extensive urban data sets, demonstrated through a case study analyzing New York City's 35.4 million 311 service requests. While ChatGPT 4.0 generated multi-modal narratives from small data samples, limitations emerged in recognizing long-term patterns across the larger data set due to constrained context windows. Overcoming these LLM limitations on complex real-world data remains an open challenge to create effective interactive community engagement agents. ACM Reference Format: