How to Get Better Results From AI: A Plain-English Prompting Guide for Small Businesses

There’s a particular kind of frustration that comes from using an AI tool; getting a response that’s technically a response to what you asked but somehow completely missing the point, and then spending ten minutes trying different variations until something useful eventually appears.

If that sounds familiar, the problem almost certainly isn’t the AI tool. It’s how you’re talking to it.

This is not a criticism. The way most people naturally ask questions, the shorthand, the context-free queries, the one-line requests all work well for search engines and for other human beings, because search engines know what you probably mean and human beings can ask follow-up questions. AI tools are neither of those things. They’re powerful, but they respond to what you actually say rather than what you mean, and understanding that one distinction makes a significant practical difference to the quality of what you get back.

This guide is about how to communicate with AI tools in a way that produces genuinely useful results. No technical knowledge required. No jargon. Just a clearer way of giving instructions that works.

A prompt is simply what you type into an AI tool. It’s your instruction, question, or request.

The same AI tool, given a well-constructed prompt, will produce something significantly more useful than it would with a vague or incomplete one. It’s not magic, it’s context. The more clearly you explain what you need, why you need it, who it’s for, and what form you want it in, the less the AI has to guess. The less guessing there is, the better the result.

A useful way to think about it is imagine you’ve hired a very capable contractor who has just started working with your business. They’re bright, knowledgeable, and genuinely helpful. But they don’t know your business, your customers, your preferences, or your tone. Every time you give them a task, you’re starting from scratch unless you provide the relevant context. Give them a clear brief and you’ll get good work. Give them a vague one and you’ll get work that technically fulfils the brief but misses what you actually needed.

That contractor is your AI tool. You are the briefer. The quality of the brief determines the quality of the output.

You don’t need to run through all five for every single interaction. Once this way of thinking becomes natural, it takes a matter of seconds. But in the early stages, knowing what the key elements are helps.

AI tools can adjust their tone, vocabulary, and approach based on the context you give them. Telling an AI “you are a helpful assistant for a small business” produces different output from “you are an experienced marketing copywriter working for a small Scottish law firm.” The second version gives the tool enough context to make sensible decisions about tone, level of formality, and approach.

One sentence establishing the context is enough.

Example: “You are a plain-English communications assistant helping a small charity write clear updates for trustees.”

The commercial AIs we use have been trained with a vast amount of knowledge, giving them a role lets them home into that specific area of knowledge.

“Write something about our new service” is a task. It’s also a task that could result in a tweet, a press release, or a 2,000-word article, and the AI will make a choice that may not match yours. “Write a 150-word paragraph for our website homepage introducing our new bookkeeping service for sole traders” is a task with enough specificity to produce something actually usable.

The more specific your task instruction, the less the AI has to guess. Think about, what format do you want? What length? What should it include? What should it avoid?

Including what not to do in your prompt is as powerful as providing what to do.

Context is everything. The AI doesn’t know your business, what you do, who your customers are, the tone you use, any constraints or requirements, all of this belongs in the prompt rather than being assumed.

If you’re asking an AI to draft a customer email, it needs to know the situation. If you’re asking it to write a policy, it needs to know who it applies to and what it’s trying to achieve. If you’re asking it to summarise a document, paste the document into the prompt.

Context in the prompt means less guessing by the AI which means better output.

Do you want bullet points or paragraphs? A table or a list? Three short options or one detailed version? Formal or conversational language? British English? A specific word limit?

If you don’t specify, the AI will choose. Sometimes it chooses well. Often it defaults to a format that needs editing before it’s useful. A brief format instruction prevents unnecessary back-and-forth.

Format instructions that make a consistent difference: “Write in plain, conversational British English.” “Use bullet points, no more than ten, each one a single sentence.” “Keep it under 200 words.” “Give me three different versions so I can choose.”

Negative instructions are just as useful as positive ones. “Don’t use technical jargon,” “don’t make it sound like a sales pitch,” “don’t use phrases like ‘in today’s fast-paced world'” (this last one turns out to be remarkably necessary) give the AI clear guardrails.

Telling an AI what not to do is often as important as telling it what to do. Include at least one constraint in any prompt where tone or content matters.

The best way to see the difference is with a real example.

The vague version: “Write a social media post about cyber security.”

What you’ll get: a generic, slightly preachy post that sounds like it was written by an insurance company. Technically a social media post. Rarely memorable.

The better version: “You are a helpful, plain-English communications assistant for a small IT consultancy based in Scotland. Write a LinkedIn post aimed at small business owners, encouraging them to check whether their staff know how to spot a suspicious email. Keep it under 100 words, conversational, not alarmist. No jargon. End with a simple question to encourage comments.”

What you’ll get: something that sounds like a person wrote it, is calibrated to the right audience, has the right tone, and is the right length. Ready to post or very nearly ready.

The second prompt took about 45 seconds to write. The time saved on editing what you get back is worth considerably more than that.

If you’re not sure what you want, ask for multiple versions and choose. “Give me three different ways to open this email” takes the same amount of time to type and gives you options instead of a single result to either accept or rework.

If the first response isn’t quite right, don’t throw it away and start from scratch. Tell the AI what needs to change. “That’s good but the tone is too formal, can you make it more conversational?” or “Can you shorten this by about half?” are effective ways to refine rather than restart.

Think of it as a conversation, not a single transaction.

If you have a complex task and you’re not sure how to frame it, try this: “I want to [describe your goal]. Before you start, ask me anything you need to do this well.” The AI will identify what context it’s missing and ask for it directly. Particularly useful for longer or more nuanced pieces of work.

If you have an example of the kind of output you’re after, a previous email in your tone, a sample post in your house style, including it in the prompt gives the AI a much clearer target. Just say “write something in a similar style to this” and paste the example.

“This is for small business owners who have no technical background” or “this is for a board of trustees, most of whom are non-executives with limited operational involvement” changes the level, vocabulary, and assumptions the AI makes. Always worth including.

AIs will give you incorrect information very confidently. They call this an hallucination. Us humans have another word for being confidently told incorrect information. Always question what the AI is telling you and ask it to cite its sources when it is giving you information. This will reduce the chance of hallucinations in the output.

Better prompting produces better output. It doesn’t produce perfect output.

AI tools can still be confidently wrong, can miss nuance, and can generate content that sounds right but contains inaccuracies. Anything consequential; financial, legal, medical, or customer-facing, should be reviewed carefully rather than assumed to be correct because it reads convincingly.

Prompting also won’t compensate for using the wrong tool for the task. And AI-generated content that goes out without a human voice applied to it tends to read that way. The prompting guide above will significantly improve what you get back, but the best results usually involve a human taking the output and making it actually sound like your business.

Think of it as a time-saving starting point. Not an end point.

When you sit down to prompt an AI tool, run through this quickly:

Who should it be? (Role)

What do I actually want? (Task)

What does it need to know to do this well? (Context)

What should it look like? (Format)

What should it avoid? (Constraints)

Not every prompt needs all five. For a quick task, two or three will do. For anything more involved, all five will produce noticeably better results than starting without them.

• NCSC AI guidance
• ICO AI and data protection
• AI Tools Are Already in Your Business
• Is AI Actually Useful for Small Businesses?
• GDPR in Plain English