Blog
How to Write AI Prompts That Work for Your Business

Most business owners who try an AI assistant have the same experience: the first answer is generic, a little bland, and not quite what they needed. The usual conclusion is “AI isn’t that useful for my work.” In most cases, the problem is not the tool. It is the request.
The prompt is the one lever you actually control. You cannot change how the model was trained or how it works inside, but you decide exactly what you ask and how you ask it. This guide, based on the free course How to Be Smarter Than AI, shows you how to write prompts that get useful results on real business tasks, often on the first try.
Why vague prompts produce vague answers
A language model works by predicting the most likely next piece of text, one step at a time. When your request is short and open, the most likely answer is the average answer: something that would fit almost anyone. That is why “write a marketing email” gives you an email that could belong to any company in any industry.
The fix is simple: give the model the details only you know. Who is it for? What is the goal? What should it look like? What must it avoid? Every detail you add narrows the range of “likely” answers toward the one you actually want.
The six ingredients of a strong prompt
The course breaks a good prompt into six ingredients. You do not need all six every time, but the more important the task, the more of them you should include.
1. Role
Tell the model which perspective to take: “Act as a B2B sales consultant” or “You are an editor for a professional services firm.” A role changes tone, vocabulary and priorities. It does not make the answer correct, though. “You are a lawyer” does not turn the model into a lawyer; legal, medical and financial decisions still need a qualified human.
2. Context
Explain the situation: your business, your audience, what has already happened. “We are a small accounting firm serving restaurant owners. Several clients missed the last tax deadline” gives the model something real to work with.
3. Instructions
State the task clearly, with an action verb: draft, summarise, compare, list, rewrite. If there are several steps, number them.
4. Format
Say what the output should look like: a table, five bullet points, a 150-word email, a subject line plus body. Format is where many prompts save the most correction time.
5. Boundaries
Say what to avoid or limit: “No jargon,” “Don’t invent figures,” “Keep it under 120 words,” “If you’re unsure, say so.” Boundaries protect you from padding and from made-up details.
6. Examples
If you have a standard in mind, show it. One good sample of a past email or report communicates tone and structure faster than a paragraph of description. Start without examples; add one or a few only when the output keeps missing your standard.
A before-and-after example
Before: “Write a follow-up email to a client.”
After: “Act as an account manager at a small digital marketing agency. A client attended our proposal meeting last week but hasn’t replied. Write a short follow-up email (under 120 words) that thanks them, restates the one main benefit we discussed (more qualified leads from their website), and suggests a 15-minute call this week. Friendly but professional tone. No pressure tactics. Give me the subject line and the body.”
The second version takes a minute longer to write. In return, you usually get something you can send after light editing instead of something you have to rewrite.
Let the AI help you write the prompt
If writing a full prompt feels like work, use a shortcut the course calls “let it write the prompt for you.” Describe your task and ask: “Write me the best prompt for this task using role, context, instructions, format, boundaries and examples. Don’t do the task yet.” Then review the draft, correct anything that is wrong or missing, and run it in a new, clean chat. The trick saves time, but it never replaces your review.
Practical checklist before you hit Enter
- Is the goal clear in one sentence?
- Have I given the context only I know (audience, situation, constraints)?
- Did I use a clear action verb for the task?
- Did I specify the format and length?
- Did I state what to avoid, including “don’t invent facts”?
- Would one example help the model match my standard?
- Did I put the most important requirement first, not buried in the middle?
- Is there any sentence I could delete without changing the result? If so, delete it.
Common mistakes
- Contradictory instructions. “Short but very detailed” forces the model to guess. Decide what matters more and say it.
- Padding. Long polite introductions, repeated instructions and irrelevant background cost time and usage without improving the answer. The shortest prompt that is enough is usually the best one.
- Treating a role as expertise. A role shapes the style of the answer, not its accuracy. Verify anything that matters.
- Burying the key condition. Models pay less attention to details lost in the middle of long text. Lead with what matters most.
- Pasting sensitive data to “give context.” Use placeholders such as [client name] instead of real names, numbers or confidential details.
- Accepting the first draft as final. Treat output as a draft from a fast, capable intern. Review before it goes to a client.
Frequently asked questions
Do I need to use all six ingredients every time?
No. For quick, low-stakes tasks, a clear instruction and format may be enough. For anything client-facing or important, the course suggests including at least four of the six.
Are longer prompts always better?
No. Length helps only when it adds information the model needs. Repetition, flattery and contradictions make results worse, not better. Test by deleting a sentence: if the result doesn’t change, it wasn’t needed.
Why do I get a different answer when I ask the same thing twice?
That is normal. Models introduce some variation by design, so wording changes between attempts. If the substance changes a lot, your prompt probably leaves too much open, or the topic needs verification.
Should I save prompts that work well?
Yes. Keep a small library of prompts for your recurring tasks (proposals, follow-ups, summaries). Many AI tools also let you save instructions in projects or reusable skills, which the course covers in later stations.
Take the next step
Writing good prompts is a skill, and it improves quickly with practice. If you want to go deeper, the free course How to Be Smarter Than AI covers this in detail across four short stations: writing your first prompt the right way, letting AI draft the prompt for you, role prompting and prompting with examples. It is free, practical, and written for people who want results, not theory. Start the course here.