Back to All Terms
AI Definition

Prompt Engineering

The craft of designing inputs that reliably steer a language model toward accurate, useful outputs.

Prompt engineering is the practice of crafting, structuring, and iterating on the instructions given to a language model so that it produces consistent, accurate, and useful output, without changing the model's underlying weights. Because LLMs are highly sensitive to phrasing, examples, and formatting, small changes in a prompt can significantly change output quality. Prompt engineers work with techniques like few-shot examples (showing the model sample input-output pairs), chain-of-thought prompting (asking the model to reason step-by-step before answering), and system-level instructions (persistent rules that shape tone, scope, and safety behavior across an entire conversation). In production settings, prompt engineering overlaps heavily with evaluation: testing prompts against real user inputs and measuring failure modes before shipping.

Practical Implementation & Example

From a Vague Prompt to a Production-Ready One:

Weak prompt: "Summarize this."

Engineered prompt:

You are a financial analyst. Summarize the following earnings call transcript in exactly 3 bullet points, each under 20 words. Focus only on revenue, guidance, and risk factors. If a category is not mentioned, write "Not disclosed."

TRANSCRIPT:
{transcript_text}

Core Techniques Used Daily:

  1. Role Assignment: "You are a..." anchors the model's tone and domain expertise.
  2. Few-Shot Examples: Providing 2-3 sample Q&A pairs before the real question dramatically improves format consistency.
  3. Chain-of-Thought: Adding "think step-by-step before giving your final answer" improves accuracy on multi-step reasoning and math tasks.
  4. Output Constraints: Explicit length, format, and structure rules (JSON schemas, bullet counts, word limits) reduce the need for post-processing.