How to Write AI Prompts That Actually Work: A Simple 4-Part Formula

Most people try AI by typing something short and hoping for the best. "Write me an email." "Summarise this document." "Give me some ideas." The result comes back instantly — and it is almost always vague, generic, or slightly off. You read it, sigh, and either spend fifteen minutes editing it into shape or abandon it altogether. It's frustrating — especially when everyone around you seems to be getting magic results and you're left wondering what you're doing wrong.

This is not a skills problem. Nobody teaches you how to write prompts that produce genuinely useful output. (If you're not sure where AI fits into your work at all, start with finding your first AI win — then come back here for the technique.) Most AI prompt tips you will find online are either absurdly complex or so generic they amount to "be more specific." Neither is particularly helpful when you are staring at an empty chat window during a busy workday.

The good news: writing better prompts is not difficult. It does not require technical knowledge or hours of practice. It requires a simple structure — the same kind of structure you would use when briefing a capable colleague.

Why Vague Prompts Fail

Consider this prompt: "Write me an email to my team about the project."

What you get back will almost certainly be bland, overly formal, and uselessly generic. It will not sound like you, it will not mention the details that matter, and you will end up rewriting it from scratch. The AI did not fail you — you just did not give it enough to work with.

Think of it this way. If you walked up to a colleague and said "write me a thing," you would expect a confused stare, not a brilliant first draft. Your colleague would need to know who it is for, what you are trying to say, how long it should be, and what tone to use. AI needs exactly the same briefing.

The difference between a mediocre AI output and a genuinely useful one is rarely about which tool you use. It is about how clearly you describe what you need. This is one of the most practical AI prompt tips you can learn: treat every prompt as a short brief.

The Four-Part Prompt Formula

Once you see the pattern, it becomes second nature. Of all the AI prompt tips in this article, this formula is the one worth remembering. Every effective prompt contains four elements — not always in this exact order, but always present in some form. Full transparency — we didn't start with this formula. Our early prompts were just as vague as the examples above. It took dozens of underwhelming results before we noticed the pattern that actually works. This is not a rigid template. It is a way of thinking about what the AI needs from you before it can produce something worth using.

Context — who you are and why

Tell the AI what situation you are in. Your role, your audience, the purpose behind the request. "I am a marketing manager preparing a team update" gives the AI far more to work with than just asking for an email. Context shapes tone, vocabulary, and level of detail. Even a single sentence of context — who you are and who will read the output — changes the result dramatically.

Task — what you need

Be specific about the actual output you want. Not "write something about the project" but "draft a 150-word email announcing that the deadline has moved." The clearer your task, the less editing you will do afterwards.

Format — how it should look

Specify the shape of the output. Do you want bullet points or full paragraphs? A numbered list? A specific word count? Should it include headings? Format instructions prevent AI from making assumptions that create extra work for you.

Constraints — what to include or avoid

In our experience, this is where most people stop short. Constraints are the guardrails that keep the output relevant. "Focus on financial impact." "Avoid jargon." "Do not use bullet points." "Keep the tone collaborative." These small additions make a disproportionate difference to the quality of what comes back. Without constraints, AI tends to default to a generic tone and tries to cover everything — which usually means it covers nothing well.

You do not need to write a long prompt to include all four parts. Two or three sentences will often do the job. The point is not length — it is clarity. Once you know how to write AI prompts with this structure, the habit forms quickly.

🧠 Quick Challenge: You need to ask AI to prepare a short client update email. Which of these two prompts will produce a more useful first draft?

• A) "Write a client update email about our project timeline. Make it professional."
• B) "I am a project manager. Draft a 100-word email to our client explaining that the design phase is two weeks ahead of schedule. Tone: confident but not boastful. Mention that we will use the extra time for additional user testing."

Three Workplace Examples

The formula only matters if it works in practice. The best AI prompts for work are not clever or elaborate — they are clear. Here are three everyday tasks, each shown first with a vague prompt and then with an improved version. Notice how the improved prompts are not dramatically longer — they are just more specific.

Drafting an email

Vague prompt: "Write me an email to my team about the project."

Improved prompt: "I am a marketing manager. Draft a 150-word email to my team of six announcing that the website redesign deadline has moved from 15 March to 1 April. Tone: direct but reassuring. Mention that the timeline change gives us space to finish user testing properly. Do not use bullet points."

Why it works: Context (marketing manager, team of six), Task (announce deadline change), Format (150 words, no bullets), Constraints (direct but reassuring tone, mention user testing). The AI now has enough to produce something you might actually send — with light edits rather than a full rewrite. The prompt took about thirty seconds longer to write. The output saved ten minutes of rewriting.

Summarising a report

Vague prompt: "Summarise this report."

Improved prompt: "I am preparing a board briefing. Summarise this 40-page sustainability report into five bullet points, each under 30 words. Focus on financial impact and regulatory deadlines. Avoid technical terminology."

Why it works: Context (board briefing), Task (summarise into bullet points), Format (five bullets, 30-word limit), Constraints (financial and regulatory focus, no jargon). Instead of a rambling three-paragraph summary, you get a tight set of points pitched at exactly the right audience. The same report, summarised for a different audience — say a project team — would need a different prompt with different constraints. The formula stays the same; the inputs change.

Meeting preparation

Vague prompt: "Give me some points for my meeting."

Improved prompt: "I am a product manager meeting our development team tomorrow to discuss the Q2 roadmap. Generate five talking points that balance feature requests from sales with our current engineering capacity. Format as a numbered list with one sentence per point. Keep the tone collaborative, not directive."

Why it works: Context (product manager, Q2 roadmap discussion), Task (five talking points), Format (numbered list, one sentence each), Constraints (balance sales and engineering, collaborative tone). You get a focused agenda that reflects your actual situation, not a set of generic meeting advice. You will almost certainly adjust a few points — but adjusting is far quicker than starting from nothing.

In each case, the improved prompt took roughly thirty seconds more to write. The time saved in editing was several minutes — and the output was genuinely useful rather than something to endure.

Even with the four-part formula in hand, two patterns tend to trip people up. These are worth knowing because they undermine otherwise solid AI prompt tips.

Over-prompting

Not every task needs a detailed brief. If you are asking AI to brainstorm ten blog title ideas, a short prompt is perfectly fine. Writing a 500-word prompt for a simple task wastes your time and can actually confuse the output. If you've done this, you're not alone — it's a natural overcorrection when you first learn that detail matters. Match the effort of your prompt to the complexity of the task. A quick question deserves a quick prompt. A nuanced deliverable deserves a more considered one.

A useful rule of thumb: if the task would take you under five minutes to do yourself, keep your prompt to one or two sentences.

Using AI as a search engine

Many professionals default to asking AI factual questions — "What is the latest UK corporation tax rate?" or "When was this regulation introduced?" AI tools can generate plausible-sounding answers to these questions, but they are not reliable fact-checkers. They are drafting assistants.

AI is at its most useful when you ask it to produce, shape, or restructure content — not when you ask it to retrieve facts. For factual lookups, use the sources you already trust. For drafting, summarising, reformatting, and brainstorming, AI is genuinely good — and the four-part formula will help you get the most from it.

Start With One Prompt This Week

The best AI prompt tips are the ones you actually use. You do not need to memorise a system or overhaul your workflow. (Want to build these skills into a broader practice? The AI for Beginners learning path picks up where this article leaves off.) Pick one task from your week — an email, a summary, a set of talking points — and try writing a prompt that includes all four parts: context, task, format, and constraints.

Notice the difference. If the output is closer to what you need than what you usually get, you've found something worth repeating. That's how a sustainable AI practice begins — not with a grand plan, but with one task done slightly better. You've got everything you need to start. The rest is practice.

Ready-Made Prompts — Browse the AI Tutorium Prompt Library for templates you can use right now — each one built with the four-part formula.