AI-powered data experiences are rapidly evolving, with large language models (LLMs) enhancing data consumption by enabling AI agents to interpret and act on business data. However, LLMs face limitations such as hallucinations and require a semantic layer to provide context, ensuring accurate data processing. This semantic layer organizes data into business definitions, allowing LLMs to query data effectively while avoiding direct interactions with complex database schemas. Cube plays a crucial role in this ecosystem, serving as an interface atop data warehouses, simplifying the querying process, and addressing performance and security concerns through caching and access control. By combining LLMs with semantic layers, Cube facilitates the development of AI-powered applications that can efficiently manage and interpret complex data queries.