This session explores how prompt engineering can revolutionize big data preprocessing workflows by leveraging the power of foundation models like LLMs. Traditionally, data preprocessing involves custom pipelines with complex logic and high maintenance overhead. In contrast, prompt-based data transformation offers a flexible, reusable, and human-readable approach that democratizes data handling across domains. We will walk through real-world examples and experiments showcasing how intelligent prompts can clean, structure, and contextualize large datasets with minimal code and maximal adaptability. This paradigm significantly reduces engineering time and makes advanced data processing accessible to analysts, domain experts, and developers alike. Attendees will leave with a new perspective on how foundation models can serve as the core of modern big data systems.