Prompt Engineering
Prompt engineering is the practice of designing and refining the instructions, context, and examples given to a generative AI model so it produces relevant, accurate output. A well-engineered prompt states the task clearly, supplies needed background, sets the format and constraints, and is tested and revised based on results.
Common techniques include giving the model a role, providing a few worked examples, breaking a complex task into steps, specifying the output format, and including the source documents the answer should be based on. Prompts are tested against sample inputs and adjusted until results are consistent.
For example, instead of asking a model to summarize a quality report, a prompt might specify the audience, ask for the top three defect types with counts taken only from the attached report, and require the answer as a short list.
Prompt engineering differs from fine-tuning. Prompt engineering changes the input to an existing model, while fine-tuning changes the model itself by training it further on new data.