Prompt engineering is the practice of designing the instructions, context, format, and constraints that guide an AI model toward useful, reliable outputs. It includes techniques like chain-of-thought prompting, few-shot examples, system prompt design, output format specification, and negative examples.
A claims triage team that notices poor accuracy on a specific claim type, adds three well-chosen examples to the prompt, and improves accuracy from 72% to 91% — without touching the underlying model — is doing prompt engineering. The model did not change. The prompt changed. That improvement matters.
