Prompt Engineering

Generative AI models are highly literal. They do not know what you want unless you explicitly tell them. "Prompt engineering" is the skill of writing clear, highly specific instructions to get the exact output you need, reducing errors and saving you time.

The CREATE Framework

A simple prompt like "Write an email about the new IT policy" will yield a generic, likely unusable result. Instead, build a robust prompt by defining Context, Request, and Tone.

Context & Persona

Use Case: Grounding the AI so it understands its role and audience.

Example: "Act as a senior IT communications officer. The audience is university staff who are not highly technical. We are rolling out a new mandatory cybersecurity training module."

Request & Format

Use Case: Telling the AI exactly what to produce and how it should look.

Example: "Write a 200-word introductory email. Use a welcoming tone. Format the email with a clear subject line and use bullet points for the instructions."

Advanced Techniques

Once you master the basics, try these techniques to dramatically improve your results:

  • Few-Shot Prompting: Give the AI examples! Paste in two or three previous emails you've written, and say: "Analyze the tone and vocabulary of these examples. Now, write a new piece matching this exact style."
  • Chain of Thought: Add the phrase "think step-by-step" to the end of a complex prompt. This forces the model to break down its logic before answering, drastically reducing logical errors.
  • Negative Constraints: Tell the AI what you don't want. For example: "Do not use marketing jargon, do not exceed 300 words, and do not use the word 'delve'."

Iteration is Key

Treat the AI like a conversational partner. If the output isn't quite right, do not start over. Reply with a targeted correction:

  1. "Make the second paragraph less formal."
  2. "Expand on bullet point three."
  3. "Re-write this for a student audience instead of staff."

Avoiding Hallucinations

  • Always constrain the AI to provided text if summarising: "Using ONLY the text provided below..."
  • Ask the AI to cite its sources or tell you if it doesn't know the answer.
  • Never accept specific statistics, legal clauses, or dates without human verification.

Building a Prompt Library

When you find a prompt that works perfectly for a recurring task (e.g., generating a weekly report summary), save it in a Word document or OneNote file. You can create a departmental "prompt library" to share best practices with your colleagues and ensure consistent outputs across your team.

Security Reminder

No matter how good your prompt is, never include sensitive, personally identifiable information (PII) or passwords within the text of your prompt when using public generative AI tools.

These pages are currently in beta. To give feedback, please contact DTS at ai.training@reading.ac.uk.