Using AI Effectively

Essay 6 — Understanding AI Course

Here's something worth knowing: AI tools like ChatGPT, Claude, and Grok are genuinely powerful — but most people don't get nearly as much out of them as they could. The difference between a useful AI interaction and a frustrating one often comes down to how you ask. In this essay, we'll look at how to get the best from your AI, how to write effective system prompts, and crucially — what you should never do with a large language model.

Talking to AI — Why How You Ask Matters

One of the most important skills in the AI age is often called prompt engineering — the art of asking questions and giving instructions in ways that get genuinely useful answers.

The thing is, you don't need to be a tech expert to do this well. It mostly comes down to a few straightforward principles:

Be specific

The single biggest improvement most people can make is to be more specific. Compare these two prompts:

❌ "Write me an email"

✅ "Write me a professional email to my manager, asking for a meeting next week to discuss the project delays we've been having. Tone should be respectful but honest — I want to address the issue, not hide it. Keep it under 150 words."

The second one gives the AI everything it needs to produce something actually useful. The first one leaves the AI guessing.

Give context

The more context you provide, the better the response. If you're asking for help with something specific to your life, work, or situation — say so.

❌ "Give me advice about my job"

✅ "I'm a secondary school teacher who's been in the job for 20 years. Lately I've been feeling burned out and I'm wondering if I should change career. What factors should I think about?"

Context doesn't need to be enormous. Just a sentence or two of background helps enormously.

Break it down

If you have a complex request, don't try to get everything in one go. Work step by step.

❌ "Write me a complete business plan"

✅ "First, let's start with the executive summary — just the key points. Then we'll do each section one by one."

You can actually ask AI to do this: "What information would you need from me to write a good business plan? Let's gather that step by step."

Ask for the response style you want

Want it in bullet points? Plain English? Academic language? Formal or casual? Tell the AI explicitly.

✅ "Explain this to me like I'm 14 years old"
✅ "Give me three pros and three cons"
✅ "In no more than 200 words..."
✅ "Write it in the style of a newspaper article"

Ask follow-up questions

Don't be afraid to iterate. AI conversations aren't like searching the web — you can refine and develop the output over multiple turns.

"That's helpful, but can you make it less technical?"
"Good start — now can we expand section two?"
"That's not quite what I meant. Let me rephrase: ..."

Think of it as a working conversation, not a single question-and-answer transaction.

System Prompts — Giving AI a Personality or Role

Here's something powerful: you can give an AI a persistent personality, role, or set of instructions that shapes how it behaves throughout a conversation. This is done through what's called a system prompt (sometimes called a system message).

What is a system prompt?

When you open a conversation with ChatGPT or Claude, there's usually a hidden instruction telling the AI: "You are a helpful assistant." That instruction comes from the company that built the AI. But you can override it.

For example, you might start a conversation with:

"You are a sceptical journalist who always asks follow-up questions and challenges assumptions. You're writing a feature article, so you're curious and thorough."

From then on, the AI behaves differently — pushing back, asking probing questions, writing in a journalistic style.

Why is this useful?

System prompts let you tailor the AI to the task at hand:

• A legal adviser persona for checking contracts
• A tutor persona for explaining concepts
• A copywriter persona for marketing materials
• A debugger persona for coding problems
• A creative partner persona for brainstorming

Tips for writing effective system prompts

Here are some principles from what actually works:

1. Be clear about the role: Give the AI a specific persona or expertise area, not just "be helpful."

2. Describe the output format: Tell it how you want the response structured.

3. Set boundaries: What should it avoid? What shouldn't it do?

4. Give an example if possible: "For instance, like this: ..."

5. Specify the tone: Formal? Casual? Encouraging? Critical?

6. Keep it reasonable: A system prompt of twenty pages won't help — be focused and concise.

Example system prompts

For a patient learning companion:

You are a patient, encouraging learning companion for a UK GCSE student. You explain things clearly and simply, use examples from everyday life, and never make the student feel stupid for asking questions. You celebrate small wins. You adapt your explanations based on what the student already knows.

For a professional editor:

You are an experienced editor reviewing a draft. You are direct but kind. You identify the three most important things that need improving, and explain why each matters. You also identify what's working well. You never rewrite the text yourself — you show the writer what needs to change and why, and let them make the changes.

Things You Should Never Do with a LLM

Now for the important part. Using AI effectively means knowing its limits. But it also means knowing what you should never do with it — because it could harm people, break the law, or undermine trust in ways that matter.

1. Never share genuinely private, sensitive information

Whatever you type into a corporate AI could be stored, reviewed, and used to train future models. Don't share:

• National Insurance numbers or passport details
• Medical records
• Financial information
• Other people's personal data without their consent
• Passwords, security credentials, or trade secrets

This includes information about other people — not just yourself. Sharing someone else's private information without their consent is potentially a data protection violation in the UK.

2. Never use AI to impersonate a real person

This includes:

• Creating fake reviews, testimonials, or endorsements
• Generating content that appears to come from a real individual
• impersonating a public figure or official body to deceive people

This is potentially fraudulent, and in some contexts could be illegal (see the Online Safety Act 2023 and fraud-related legislation in the UK).

3. Never use AI to generate harmful content

This should go without saying, but: don't use AI to create content designed to:

• Harm, threaten, or harass individuals
• Spread misinformation about real people or events
• Generate sexual imagery of real people (especially minors)
• Incite violence or terrorism

4. Never use AI to make high-stakes decisions alone

AI should assist human judgment, not replace it — especially in high-stakes situations. Don't rely solely on AI for:

• Medical diagnoses or treatment decisions
• Legal advice
• Financial decisions with significant consequences
• Decisions about people's lives, jobs, or opportunities (without human review)

5. Never pass off AI-generated content as entirely your own original work in academic or professional contexts

Most institutions now have policies on AI use. Academic dishonesty — including presenting AI-generated work as your own — can result in serious consequences. Be honest about what you've used AI for, and follow your institution's guidelines.

6. Never rely on AI as your only source for important factual information

We covered hallucinations in Essay 2, but it bears repeating: AI can generate confident-sounding false information. On matters of fact — medical, legal, financial, historical — always verify what AI tells you with authoritative sources.

7. Never use AI to circumvent safeguarding or safety procedures

If your organisation has safeguarding or safety procedures, AI cannot be used to shortcut them. For example: an AI cannot replace a mandatory DBS check, a required training module, or a human risk assessment. These exist for good reasons.

A Practical Note on Ethical Use

The reason these rules exist isn't to limit you — it's to protect you and the people around you. AI is a tool. Like any tool, it can be used well or badly. Using it well means:

• Being honest about what it is and what it can and can't do
• Taking responsibility for what you use it to produce
• Thinking about the impact of your AI use on other people
• Using it to augment your own thinking, not replace it

One useful question to ask yourself: "Would I be comfortable if my AI use was public and fully transparent?" If the answer is no, that's probably a sign to reconsider.

What Have We Learned?

• Being specific, providing context, and asking for a particular format dramatically improves AI responses
• System prompts let you tailor AI's behaviour to a specific role or task — worth experimenting with
• Knowing what not to do with AI is as important as knowing how to use it
• Never share private data, impersonate people, use AI for high-stakes decisions alone, or pass off AI work as entirely your own without disclosure

Glossary of Terms

Further Thinking

• What's the most useful system prompt you could write for your own work or life?
• Can you think of a situation where AI was used badly? What was wrong with it, and what should have been different?
• Should there be laws about how AI can be used? Who should decide?

Essay 6 of 8 — Understanding AI Course
Authors: Bea Groves-McDaniel and SAL-9000
Licensed under Creative Commons (NC, ND) — Share with attribution, no commercial use.