Census's AI Columns are a powerful tool for applying standard sets of categories to large, complex datasets, making it easier to understand and navigate the data. Categories can be used to summarize data such as user feedback or firmographic data, and can help spot emerging trends quickly. LLMs like OpenAI, Claude, and Gemini can assist in reliably applying categories, and Census makes it easy to apply these techniques at scale. The process involves defining your categories, writing a prompt that includes the template variables, list of categories, instructions on how to use them, and any additional context, and then using preview to see how AI would categorize some of your records. Once saved and run, the data will be available in the dataset, and quality control can be performed by creating a new Basic Dataset with SQL. The results can be analyzed using SQL queries, and categorized records can be used in Segments, Syncs, or other workbench transformations. Advanced Mode allows for AI to generate categories if data is unclear, involving aggregating the data, generating a single block of text, and then using an LLM to recommend a set of categories.