Prompt engineering for business users

Introduction

"I tried using ChatGPT to write a report, but it was completely off." That is a comment we often hear at Hyperbolic. AI tools like ChatGPT, Claude, and others are incredibly powerful, but like any tool, they require knowing how to use them.

Prompt engineering sounds technical and complicated, but it is actually a skill that all business users can learn. In this article, we share the techniques we use at Hyperbolic to get maximum value out of AI assistants, without needing to be a programmer or data scientist.

What is prompt engineering?

Prompt engineering is the art of formulating your questions and instructions to AI in a way that yields the best results. It's about understanding how AI models "think" and using that understanding to communicate more effectively.

Think of it as the difference between giving vague instructions to a new employee ("create a report") versus clear, detailed instructions ("create a 5-page report on Q3 sales, focusing on the North region, include graphs, and compare with last year").

Why is it important?

The difference between a good and a poor prompt can be dramatic:

Poor prompt: "Write about customer service" Result: A generic, superficial text with no value

Good prompt: "Write a 500-word article on how AI can improve customer service in retail. Focus on concrete use cases, include both advantages and challenges, and maintain a professional but accessible tone." Result: A focused, relevant article that can actually be used

At Hyperbolic, we have seen productivity increases of 50-80% among employees who learn good prompt engineering.

The four fundamental principles

1. Be specific

AI models are good at following instructions, but they cannot read minds. The more specific you are, the better the result.

Vague: "Write an email to a customer"

Specific: "Write an email to a customer who has complained about delayed delivery. Apologize for the delay, explain that it is due to weather conditions outside our control, offer a 20% discount on the next order, and keep the tone professional but warm."

2. Provide context

AI performs better when it understands the context.

Without context: "What should I reply?"

With context: "I am a project manager at a software company. A client has requested to add three new features mid-project. This will delay the delivery by two months. How can I professionally explain the consequences and suggest alternatives?"

3. Define the format

Should the output be a list, a table, an email, a note? Be explicit.

Unclear format: "Give me info about the competitors"

Clear format: "Make a table of our three largest competitors. Columns: Company Name, Market Share, Strengths, Weaknesses, Price Point. Include sources."

4. Iterate and improve

The first prompt is rarely perfect. Be ready to adjust and refine.

If the output is not good enough, ask yourself:

• Was I specific enough?

• Did I provide enough context?

• Did I clearly define the format?

• Is there missing information?

Advanced techniques for business users

Technique 1: Role-based prompting

Ask the AI to take on a specific role or perspective.

Example: "Act as an experienced CFO. Review this budget proposal and identify potential risks and cost-saving opportunities."

This prompts the AI to adopt a specific perspective and provide more targeted feedback.

Technique 2: Step-by-step instructions

For complex tasks, break the process down into steps.

Example: "I want to analyze our customer satisfaction data. Follow these steps:

1. Summarize the three key trends in the data

2. Identify areas where we are performing poorly

3. Suggest three concrete improvements

4. Prioritize the suggestions based on potential impact"

Technique 3: Examples in the prompt

Show the AI examples of what you want.

Example: "Write product descriptions in this style:

Example 1: 'This ergonomic office chair combines comfort with style. The adjustable lumbar support ensures optimal posture during long workdays.'

Example 2: 'Our wireless mouse provides precise control without clutter. 12 months of battery life means fewer interruptions.'

Now write a similar description for: [Your product]"

Technique 4: Limitations and guardrails

Be explicit about what the AI should NOT do.

Example: "Write a marketing email. IMPORTANT: DO NOT use overused sales terms like 'unique opportunity' or 'exclusive offer'. Avoid exclamation points. Keep the tone professional and fact-based."

Use cases in everyday life

Let's look at concrete ways business users can use AI in their work:

Email communication

Prompt: "Write an email to my team about changes to the meeting time. The new time is Tuesday at 2 PM instead of Monday at 10 AM. Reason: Conflict with client meeting. Tone should be friendly but professional. Apologize for the inconvenience."

Meeting preparation

Prompt: "I have a meeting with a potential customer in the construction industry. They have expressed interest in our project management software. Create an agenda for a 45-minute meeting covering: Presentation of our solution, their specific needs, demo, pricing, and next steps. Include suggested time frames for each item."

Report writing

Prompt: "Based on these key figures [insert data], write a 2-page executive summary of Q3 performance. Include:

• Brief overview

• Key results (positive and negative)

• Comparison with targets

• Recommendations for Q4 Keep the tone professional and data-driven."

Brainstorming

Prompt: "Our company needs to improve employee engagement. Current score is 6.5/10. Budget is limited. Give me 10 creative, cost-effective ideas to improve engagement. For each idea, include estimated cost and potential impact."

Data analysis

Prompt: "Here is customer data from last quarter [insert data]. Analyze it and find:

1. Trends in buying behavior

2. Which products perform best

3. Customer segments with the highest value

4. Recommendations for marketing focus next quarter Present findings in bullet points."

Working with complex documents

Summary

Prompt: "Read this 30-page contract and give me:

1. A 5-point summary of key obligations

2. Critical deadlines

3. Potential risks or ambiguities

4. Points requiring legal review Focus on business implications, not legal jargon."

Translation of technical language

Prompt: "Our IT department has sent this technical explanation [insert text]. Translate it into understandable business language for a non-technical executive. Focus on business implications and avoid jargon."

Common mistakes and how to avoid them

Mistake 1: Being too concise

Poor: "Create a marketing plan"

Better: "Create a marketing plan for the launch of our new product XYZ. Target audience is 25-40 year old professionals. Budget 100,000 DKK. Focus on digital channels. Include timeline, channels, budget allocation, and KPIs."

Mistake 2: Assuming knowledge

AI does not know your specific business, industry, or situation unless you tell it.

Poor: "How do we solve the problem?"

Better: "Our SaaS company has high churn after the trial period (40%). Customers cite complexity as the main reason. How can we improve the onboarding experience to reduce churn?"

Mistake 3: Accepting the first output

The first version is rarely perfect. Be ready to refine.

Approach: If the output is not perfect, follow up with: "This is good, but can you adjust the tone to be more [formal/informal/technical, etc.]?" or "Good start, but can you focus more on [specific area]?"

Mistake 4: Forgetting to verify

AI can make mistakes or "hallucinate" facts. Always verify important facts, figures, and claims.

Best practice: Ask for sources or verifiable information. Example: "Give me statistics about the market size for AI software in Denmark. Please include sources."

Ethical considerations

Confidentiality

NEVER share confidential or sensitive data in public AI tools. This includes:

• Customer data

• Financial figures

• Personal information

• Trade secrets

At Hyperbolic, we use enterprise versions of AI tools with data handling assurances when working with sensitive information.

Transparency

Be transparent about AI usage when relevant. If a customer receives AI-generated content, consider whether it should be mentioned.

Quality control

AI should supplement, not replace, human judgment. Always use AI as a starting point, but add your expertise and context.

Tools and resources

Popular AI assistants:

• ChatGPT: Broad applicability, good for text and analysis

• Claude: Excellent for long documents and nuanced text

• Microsoft Copilot: Integrated with the Office suite

• Google Gemini: Good for research and fact-based work

Tips for getting started

Week 1: Experiment Spend 30 minutes a day experimenting with AI on small tasks. Write emails, summarize articles, brainstorm ideas.

Week 2: Document Note which prompts work well. Build your own "prompt library" with templates for common tasks.

Week 3: Integrate Identify 2-3 daily tasks where AI can save you time. Make it a habit to use AI for these.

Week 4: Optimize Review your prompts and see where you can improve. Learn from what works and what doesn’t.

Conclusion

Prompt engineering is not rocket science, but it is a skill that requires practice. The business users who master this skill will gain a tremendous productivity boost and can focus their time on what truly requires human expertise and judgment.

At Hyperbolic, we see AI assistants as co-pilots, not auto-pilots. They are powerful tools that can significantly accelerate your work, but they do not replace the need for your knowledge, experience, and judgment.

Start small, experiment, and gradually build your prompt engineering skills. Over time, it will feel as natural as searching on Google.

Want to learn more about how your team can use AI effectively? Contact us at Hyperbolic for a workshop in prompt engineering tailored to your specific business needs.