Prompting Guide for Business Owners: How to Get the Most Out of AI Without Being Technical

You have ChatGPT, Claude, or Gemini open in your browser. You type, "write me an email for a client." The result is generic, cold, and sounds like a robot. You delete it, try again, and by the third attempt, you end up writing it yourself. Conclusion: "AI isn't useful for my business."

The problem isn't the tool. The problem is how you talk to it.

According to McKinsey data, companies that use AI effectively report productivity gains of 26% to 55%. But the difference between "using AI" and "using AI well" almost always comes down to one thing: the quality of the prompt. That is, the instructions you give it.

A vague prompt produces a generic response. A well-constructed prompt produces a result that you can use directly, without editing, in less time than it would take you to write it yourself.

This guide teaches you 7 practical prompting techniques, with real examples of tasks you do every day in your company. No technical jargon. No academic theory. Just what works.

Why Most Business Owners Get Mediocre Results

There is a pattern that repeats. The first time someone tries AI, they write something like "help me with my company's marketing" or "write a professional email." The AI responds with something correct but totally generic. The business owner is disappointed and concludes that the tool is inadequate.

But the problem is not capability. It is communication. Imagine you hire a brilliant consultant, seat them in a chair, and tell them, "improve my business" without giving them more context. They don't know what your company does, who your client is, what your specific problem is, or what you expect from them. Their response will be forcedly generic.

The same thing happens with AI. The more context you give it, the better the result. And that is exactly what "knowing how to prompt" means: giving clear instructions with defined context, format, and objective.

The most common mistakes business owners make when using AI are:

• Being too vague: "write a commercial proposal" instead of explaining who it's for, what service, and in what context.
• Not providing context: jumping straight to the request without explaining the situation.
• Putting everything in one prompt: asking the AI in one message to write an email, analyze some data, and create a marketing plan.
• Accepting the first answer: not iterating or asking for adjustments, as if the AI only had one chance.
• Not specifying format: letting the AI decide whether to respond with three paragraphs or a list of twenty points.

The good news is that correcting these mistakes is simple. The techniques that follow are learned in minutes and applied immediately.

Technique 1: Context + Task + Format

This is the most basic and most effective structure. Instead of throwing out a loose request, structure your prompt in three parts: who you are and what the situation is, what you need exactly, and what format you want it in.

Bad prompt:

Write an email for a client.

Good prompt:

I am the commercial director of a renovation company in Valencia. A client requested a quote a week ago to renovate their kitchen, and we haven't heard back. I need a follow-up email that is cordial, reminds them of our offer without pressuring them, and invites them to resolve any doubts by phone. Maximum 150 words, friendly and professional tone.

The difference is brutal. The first prompt produces an email that could come from any company to any client. The second produces an email you can send as is.

Why it works: The AI doesn't have to guess anything. It knows who you are, what the situation is, what you want to achieve, what tone to use, and how long the result should be.

Technique 2: Assigning a Role

When you tell the AI, "act as an expert in X," you change its perspective. It's not a magic technique that makes it smarter, but it guides its responses toward the type of language, depth, and focus that professional profile would use. It is especially useful for writing, analysis, and strategy tasks.

Bad prompt:

Give me ideas to improve my online store's customer service.

Good prompt:

Act as a customer experience consultant with 15 years of experience in e-commerce. My online store of gourmet products receives 40 orders daily, and the main complaints are shipping delays and lack of information about the order status. Provide 5 concrete improvements, prioritized by impact, explaining the approximate cost and implementation difficulty for each one.

The role of "consultant with 15 years of experience" makes the answers more specific, more structured, and with a level of detail that a generic prompt cannot achieve.

Technique 3: Examples (Few-Shot Prompting)

Instead of explaining what you want with words, you show it. Giving the AI two or three examples of the desired outcome is one of the most reliable techniques for achieving consistency and quality. Technicians call it "few-shot prompting," but in practice, it is simply teaching by example.

Bad prompt:

Write responses to my clients' reviews on Google.

Good prompt:

I need to respond to customer reviews on Google for my restaurant "La Terraza de Ana" in Seville. Here are two examples of the tone and style I want:

Review: "The paella was delicious, we will definitely be back" Response: "Thank you for your kind words; we are so glad you enjoyed the paella. We hope to see you again; next time, try the black rice which is also worth it."

Review: "They took a long time to serve, although the food was good" Response: "We apologize for the wait; we had an unusual peak in reservations that day. We are reinforcing the dining room staff so this doesn't happen again. Thank you for your patience and for valuing the food; we hope to see you again."

Now respond to this review with the same tone: "We went for dinner on Saturday, the place is nice but the desserts were just okay"

With two examples, the AI grasps the pattern: friendly, grateful, acknowledging what can be improved without being defensive, and with an invitation to return. Without the examples, the response would be generic and probably too formal.

Technique 4: Step-by-Step (Chain of Thought)

When you ask the AI to solve something complex—analyze data, make a decision, evaluate options—the results improve significantly if you ask it to reason step-by-step before giving the final answer.

Bad prompt:

Look at these sales data and tell me what's happening: January 45,000, February 38,000, March 52,000, April 31,000, May 29,000, June 48,000.

Good prompt:

Analyze this monthly sales data for my office supply company. I want you to follow these steps:

1. Identify the general trend (growth, decline, stable, seasonal).
2. Detect anomalous months and provide possible explanations.
3. Compare the data with typical seasonal patterns for the sector.
4. Suggest 3 concrete actions based on the analysis.

Data: January €45,000, February €38,000, March €52,000, April €31,000, May €29,000, June €48,000.

Present the analysis in an executive format that I can share with my partner.

By forcing a step-by-step analysis, you prevent the AI from jumping straight to superficial conclusions. You force it to reason, and that intermediate reasoning improves the quality of the final conclusion.

This technique is especially powerful for financial analysis, vendor evaluation, contract review, and any decision where you need the AI to consider multiple factors before giving an opinion.

Technique 5: Limiting and Restricting

Sometimes the most useful thing is not telling the AI what you want, but what you don't want. Restrictions narrow the result and prevent generic or overly long answers.

Bad prompt:

Write a commercial proposal for a new client.

Good prompt:

Write a commercial proposal for a dental clinic in Madrid that wants to implement an automatic reminder system for its patients. The proposal must:

• Not exceed one page (max 400 words).
• Start directly with the problem we solve, without generic introductions.
• Include three quantifiable benefits (reduction in missed appointments, time savings, etc.).
• End with a clear next step (15-minute call).
• Not use words like "innovative," "leader," "cutting-edge," or empty superlatives.
• Direct and professional tone, without being aggressively commercial.

Restrictions are like guardrails on a mountain road. They don't limit where you can go, but they prevent you from getting off the path.

Technique 6: Iterating, Not Starting from Scratch

AI is not a Google search engine where you type, read, and close the tab. It is a conversation. And like any conversation, you can ask for clarifications, adjustments, and changes to what it has already given you.

Many business owners treat every AI response as definitive. If they don't like it, they delete everything and start with a new prompt from scratch. That is like asking an employee to redo an entire report because the introduction isn't convincing.

Instead of starting from scratch, iterate:

• "The tone is too formal, make it friendlier."
• "Shorten the introduction and get straight to the point."
• "Add a concrete example in the second paragraph."
• "Change the focus: instead of talking about us, center the message on the benefit for the client."
• "Keep the structure but rewrite it with more specific data."

Every iteration improves the result without losing the previous work. Usually, with two or three adjustments, you reach a result that would take you fifteen minutes to write yourself.

Practical tip: If the AI's first response is not even remotely close to what you wanted, the problem is usually in your initial prompt, not the AI. Check if you gave enough context before iterating.

Technique 7: Asking It to Critique You

This is the technique fewest people use, but it provides the most value in business decisions. Instead of asking the AI to help you do something, ask it to critique what you have already done or to find flaws in your reasoning.

Bad prompt:

I'm thinking of opening a second store in another city. What do you think?

Good prompt:

I have a bicycle store in Zaragoza that generates €320,000 annually with a net margin of 12%. I am considering opening a second store in Huesca. The rent would be €1,800 per month, and I would need to hire two people.

Act as a skeptical financial analyst. Your job is to find the 5 main risks of this expansion, the hidden costs I probably aren't considering, and the realistic worst-case scenario. Do not try to validate my idea: look for the reasons why I might fail.

This technique is tremendously useful because the AI, by default, tends to be accommodating. If you ask it to validate your idea, it will validate it. If you explicitly ask it to question it, you will get a much more honest and useful analysis.

It works equally well for reviewing contracts, evaluating vendor proposals, analyzing marketing campaigns, or detecting weaknesses in your business plan.

Quick Application: 4 Daily Tasks with Ready-to-Use Prompts

To help you apply this today, here are prompts ready for four tasks that any business owner regularly performs:

Responding to an Unsatisfied Client

I am the customer service manager for [your company]. A client wrote complaining about [specific problem]. I want a response that acknowledges the problem without excuses, explains what we are going to do to solve it, and offers [specific compensation]. Empathetic but professional tone. Max 120 words.

Summarizing a Long Document

Summarize this document in 5 key points, prioritized by relevance for a SME director's decision-making. For each point, indicate the practical implication for my business and if it requires any action on my part. [paste document]

Preparing for a Meeting

I have a meeting tomorrow with [person/company] about [topic]. My goal is [what you want to achieve]. Prepare a 30-minute agenda with the key points to cover, the questions I should ask, and the data I should have prepared. Also include 3 possible objections they might raise and how to respond to each one.

Creating Social Media Content

I need 5 LinkedIn posts about [your sector topic]. I am [your role] at [your company]. My audience is [client type]. Each post must have an initial hook sentence, 3-4 lines of content with a practical data point or insight, and a question at the end to generate conversation. Do not use generic hashtags or emojis.

A Note on Different Tools

ChatGPT, Claude, Gemini, and other models have distinct personalities, but the prompting techniques are universal. A well-structured prompt works well on any tool. That said, there are nuances:

• ChatGPT is versatile and responds well to highly structured instructions. It works especially well when you give it clear restrictions on format and length.
• Claude excels at tasks requiring long and careful reasoning, such as analyzing complex documents or drafting extensive texts with coherence.
• Gemini has the advantage of being integrated with Google Workspace, making it more practical if you already work with Gmail, Docs, and Sheets.

The technique is the same. You choose the tool based on your workflow. The important thing is that learning how to write good prompts is a transferable skill: what you learn with one tool serves you with all the others.

Manual Prompting vs. Automation: When to Make the Leap

Learning to write good prompts is the first step to leveraging AI in your company. But if you find yourself repeating the same type of prompt dozens of times a day—responding to similar emails, generating the same type of report, answering the same client questions—then the next step is not to improve your prompts. It is to automate the process.

A good prompt solves a specific task. An automated system solves that same task a hundred times a day without you having to intervene.

For example:

• If you write prompts to respond to client emails, an AI chatbot on WhatsApp can do it automatically 24 hours a day.
• If you use AI to summarize documents or search for information in your manuals, RAG allows the AI to consult your documentation without you having to paste it into every conversation.
• If you repeat complex tasks involving multiple steps, AI agents can execute those entire workflows autonomously.

Manual prompting is the foundation. Automation is what scales. And between the two lies the difference between using AI as a curiosity and using it as a real competitive advantage.

How We Can Help

At Navel Digital, we help companies go beyond manual prompting. We design and implement AI systems that automate the repetitive tasks of your business: from chatbots that assist your clients with your company's real information, to agents that manage emails, generate proposals, and coordinate internal processes without manual intervention.

If every day you copy and paste the same type of instructions into ChatGPT, that task can probably be automated. Contact us at no obligation, and we will explain how.