AI Prompt Engineering

Foundations

Prompt Engineering

The practice of designing and refining inputs to generative AI so it produces accurate, relevant, and useful outputs.

Prompt

The instruction or input given to an AI model that guides what it generates.

Large Language Model (LLM)

A generative AI trained on vast text to understand and produce human-like language in response to prompts.

Why Prompting Matters

The quality of an AI's output depends heavily on the prompt; clear, well-structured prompts produce far better results.

Non-Determinism

The same prompt can yield different outputs; prompting reduces variability but does not eliminate it.

Techniques

Be Clear and Specific

Stating exactly what you want, in detail, reduces ambiguity and improves output quality.

Provide Context

Giving relevant background, audience, and constraints helps the model tailor its response.

Assign a Role

Telling the AI to act as a specific persona (e.g., 'act as an editor') shapes tone and expertise.

Specify Format

Asking for a particular structure — list, table, steps, length — makes output easier to use.

Few-Shot Prompting

Including examples of desired input-output pairs to guide the model's response.

Zero-Shot Prompting

Asking the model to perform a task with instructions but no examples.

Chain-of-Thought

Prompting the model to reason step by step, which improves results on complex tasks.

Constraints and Guardrails

Setting boundaries — what to include, avoid, or how long — to keep output on target.

Refining

Iteration

Refining a prompt based on the output, adjusting wording, context, and constraints until results improve.

Decomposition

Breaking a complex request into smaller prompts or steps for better, more controllable results.

Prompt Templates

Reusable prompt structures that produce consistent results for recurring tasks.

Testing Prompts

Trying variations and comparing outputs to find what works reliably.

Quality

Verifying Output

Always checking AI output for accuracy, bias, and appropriateness before use, since models can hallucinate.

Hallucination

When AI confidently produces false or fabricated content; prompts and verification reduce its impact.

Responsible Use

Protecting Data

Avoiding entering sensitive or proprietary information into AI tools that may expose it.

Bias Awareness

Recognizing that outputs can reflect bias in training data and reviewing accordingly.

Human in the Loop

Keeping people accountable for reviewing and approving AI output, especially for important decisions.