How to Write AI Prompts That Get Useful Answers: A Practical Guide for Small Businesses
Most bad AI output comes from vague prompts. Here is a simple five-part structure, real before-and-after examples, and the mistakes to avoid.
If you have tried an AI assistant and got a bland, generic answer, the problem was probably not the tool. It was the prompt. AI models respond to exactly what you give them. Give them a vague request and they fill the gaps with the most average possible answer.
The good news is that writing better prompts is a learnable skill, and it takes about ten minutes to get noticeably better results. This guide gives you a simple structure you can reuse for almost any business task, with real before-and-after examples.
In this article
Why vague prompts give vague answers
An AI model does not know your business, your customers, your tone of voice or what you plan to do with its answer. When you type "write a product description for my candles", it has to guess all of that. It guesses the safe middle: generic adjectives, no specifics, a length that suits nobody in particular.
Think of the AI as a capable new assistant on their first day. They are smart and fast, but they know nothing about your situation. The more context you give, the less they have to guess.
The five-part prompt structure
You do not need every part every time, but running through this list before you hit enter fixes most weak prompts.
1. Role and context
Tell the AI who it is helping and what the situation is. "I run a small handmade candle shop that sells online to customers in the UK" is far more useful than nothing.
2. The task
Say exactly what you want produced. Use a verb: write, summarise, compare, rewrite, list, critique. One clear task per prompt usually works better than three bundled together.
3. The audience
Who will read the output? A busy customer on a phone, a bank manager, a technical buyer? The same information reads very differently for each.
4. Constraints
Length, tone, words to avoid, facts that must be included, reading level. Constraints are where you turn generic text into your text.
5. Output format
Do you want a table, bullet points, an email with a subject line, three options to choose from? Say so. If you do not, you will get whatever the model defaults to.
Before and after: three real examples
Product description
Before: Write a product description for a lavender candle.
After:
I run a small UK candle shop. Write a product description for our
lavender and cedar soy candle (200g, 40-hour burn, cotton wick,
hand-poured in Bristol).
Audience: shoppers browsing on mobile who want a calming gift.
Tone: warm, simple, not flowery. Avoid the words "luxurious" and "indulge".
Length: under 80 words, then 3 short bullet points with the key facts.The second prompt produces something you can actually publish, because every fact and the tone came from you.
Customer email
Before: Write an email saying the order is late.
After: Write a short apology email to a customer whose order is 5 days late because our courier lost the parcel. We have re-sent it by tracked delivery (arrives Thursday) and refunded the shipping fee. Tone: sincere, not grovelling. Under 120 words. Include a subject line.
Summarising a document
Before: Summarise this.
After: Summarise the contract below for a small business owner with no legal background. List: (1) what we must do, (2) what the supplier must do, (3) any automatic renewals or cancellation notice periods, (4) anything unusual I should ask a lawyer about. Use plain English.
Important: an AI summary of a contract is a starting point for your own reading, not legal advice. Use it to spot what to ask a professional about.
Techniques that make a big difference
Give an example of what good looks like
If you have a past email, post or description you liked, paste it in and say "match this style". Examples communicate tone far better than adjectives like "friendly" or "professional".
Ask for options, then pick
"Give me five subject lines" is often more useful than "give me a subject line". You get to choose, and seeing several versions helps you notice what you actually want.
Iterate instead of starting over
Your first result is a draft. Reply with specific feedback: "shorter", "less formal", "the second paragraph repeats the first", "add the price". Small corrections in the same conversation are faster than rewriting your whole prompt.
Ask it to ask you questions
For bigger tasks, end your prompt with: Before you start, ask me any questions you need answered to do this well. The AI will often surface details you forgot to mention.
Break big jobs into steps
Do not ask for "a full marketing plan" in one go. Ask for an outline first, agree on it, then work through each section. You stay in control and the quality is much higher.
Common mistakes to avoid
- Trusting facts without checking. AI models can state wrong information confidently, especially numbers, dates, laws, prices and quotes. Verify anything that matters against an official source.
- Pasting sensitive data. Do not paste customer personal data, passwords or confidential financial details into a tool unless you have checked its data and privacy settings and your business is allowed to.
- Publishing unedited output. Readers and search engines value content with real experience and specifics. Use AI to draft and organise, then add what only you know: your prices, your process, your stories.
- Overloading one prompt. Five tasks in one message usually means five mediocre answers.
- Being polite but vague. "Could you maybe make it a bit better?" gives the model nothing to act on. Say what "better" means.
A reusable prompt template
Copy this into a note and fill in the blanks whenever you start a new task:
Context: I am [role] at [type of business] serving [customers/location].
Task: [verb] a [thing].
Audience: [who will read it and where].
Must include: [facts, numbers, links].
Avoid: [words, claims, topics].
Tone: [e.g. plain and friendly] — here is an example I like: [paste].
Format: [length, bullets/table/email, number of options].
If anything is unclear, ask me before you start.The bottom line
Better prompts are not about magic phrases. They are about giving the AI the same briefing you would give a new team member: who it is for, what you need, what to avoid and what the finished thing should look like. Start with the template above on your next task and compare the result with what you used to get. The difference is usually obvious on the first try.
