Implementing zero-shot text classification using modern LLMs to process millions of documents without labeled data.