Prompt Engineering

Prompt engineering is the practice of crafting the instructions you give an AI tool so that it produces the most accurate, useful, and appropriately formatted response. A prompt is simply what you type into a chatbot or AI system, but the way you phrase it has a direct effect on the quality of what comes back. Writing prompts thoughtfully is a learnable skill, and it does not require any technical knowledge to start.

When a large language model receives your input, it uses that text as its only guide to what you want. Unlike a human colleague who can ask clarifying questions or draw on shared context, the model works only with what you give it. Prompt engineering is the discipline of making that input as informative and precise as possible, so the model can do its best work.

How it works#

Every prompt is processed as a sequence of tokens, and the model predicts the most appropriate response based on all of them. A vague prompt leaves many possibilities open, so the model makes assumptions that may not match what you had in mind. A well-constructed prompt narrows the possibility space towards the output you actually need.

Effective prompts typically include some combination of the following:

• Context: who you are, what the situation is, or what the output is for.
• Task: a clear description of what you want the model to do.
• Format: the structure you want the output to follow (bullet list, table, short paragraph, and so on).
• Constraints: things to include or avoid, the reading level, the word count, the language variant (UK English, for example).
• Examples: a short sample of the style or approach you are after.

You do not need all of these in every prompt. Even adding one element, such as specifying the audience or the desired length, tends to improve results noticeably.

Examples#

Weak vs strong prompt for a CV review. Instead of typing "help with my CV", a better prompt reads: "Review my CV for a marketing manager role at a mid-sized UK retailer. Suggest three specific improvements to the work experience section. Use UK English and keep your feedback under 200 words." The second version gives the model the role, the audience, the scope, the format, and a constraint. The response will be far more actionable.

Using a persona instruction. Adding "Explain this as if I have no background in finance" to a question about pension contributions will produce a clearer, jargon-free answer. Asking the model to act as a critical editor rather than a helpful assistant produces more candid feedback. These small additions cost a few extra words but change the character of the response entirely.

Why it matters#

As AI tools become part of everyday work, the gap between those who know how to use them well and those who do not is becoming visible in output quality and speed. A poorly written prompt can produce generic, off-target content that needs heavy editing. A well-written prompt can produce something close to final-draft quality at the first attempt.

The good news is that prompt engineering is not gatekept by any technical barrier. Anyone who can write a clear sentence can learn it. The habits involved, being specific, providing context, stating what you want, are the same habits that make any written communication effective.

Explore the Write Better Prompts tutorial for practical, step-by-step guidance, or browse prompt templates for ready-to-use starting points you can adapt to your own work.