Netflix is enhancing its Graph Search platform by integrating AI to simplify user queries. Traditionally, users needed to learn complex structured query languages to filter data. For instance, a user wanting to find movies from the 90s about robots had to navigate through various UI elements that converted their input into a valid query. This involved cumbersome processes that varied across applications, making it hard for users to achieve their goals efficiently. The introduction of Large Language Models (LLMs) aims to streamline this process. The article explains how Netflix's system converts ambiguous natural language queries into structured Graph Search Filter DSL statements. Achieving this involves ensuring three types of correctness: syntactic (the query must parse correctly), semantic (the query must respect the data types and available fields), and pragmatic (the query must capture the user's intent). To facilitate the LLM's task, Netflix prepares context by providing metadata and controlled vocabularies from GraphQL schemas. Controlled vocabularies define a limited set of acceptable values for specific fields, ensuring that generated queries remain accurate. By focusing on this AI-driven approach, Netflix is working towards an intuitive search experience that reduces the need for users to learn complex query structures.