How to Talk to AI: A Beginner's Guide to Prompting

Florian Tisson

Florian is one of five co-founder from Cobey AI and a qualified entrepreneurship student. At Cobey AI, he characterizes the role of CFO as a way to manage finances sustainably.

How to Talk to AI: A Beginner's Guide to Prompting

Learn the simple art of getting better results from AI tools

Have you ever asked ChatGPT a question and gotten a response that was completely off-base? Or wondered why some people seem to get amazing results from AI while yours are just... okay?

The secret is prompting, the art of writing clear instructions for AI tools. Just like talking to a helpful but very literal assistant, the way you communicate makes all the difference.

What is Prompting?

Prompting is simply giving clear instructions to AI using everyday language. Instead of typing code like traditional programming, you have a conversation with the AI to tell it what you want.

Key insight: AI doesn't actually "understand" like humans do—it's incredibly good at recognizing patterns and generating text that fits those patterns. This means:

  • Clear patterns work better than vague requests
  • Context helps the AI choose the right response style
  • Specificity prevents generic, unhelpful answers

Think of AI as an extremely knowledgeable assistant who takes instructions very literally. The clearer your instructions, the better your results.

The Three Ways AI Learns From You

Zero-Shot: Just Ask

Simply ask the AI to do something without examples.

When to use: Common, straightforward tasks Example: "Write a professional email declining a meeting invitation"

One-Shot: Show One Example

Give the AI one example to follow for format and style.

Example: "Write product descriptions following this format: Example: 'Wireless Headphones - Experience crystal-clear sound with 20-hour battery life. Perfect for commuters and music lovers. $99.99'

Now write one for: Smart Water Bottle"

Few-Shot: Show Multiple Examples

Provide several examples for the most consistent results.

Example: "Create social media captions like these:

  • ' Monday motivation: Fresh coffee, fresh goals! What's one thing you're excited about this week?'
  • ' Wellness tip: Take three deep breaths before checking emails. Your stress will thank you!'
  • ' Friday learning: One new skill can change everything. What could you learn in 15 minutes today?'

Now create captions about weekend planning."

What the AI learns: Emoji use, question format, positive tone, hashtag style

Step-by-Step Thinking: The Game Changer

Add "Let's think step by step" to your prompts for dramatically better results. Research shows this simple phrase can improve AI accuracy from 18% to 78% on complex problems!

Why it works:

  • Errors become visible and fixable
  • Complex problems get broken into manageable pieces
  • The AI makes fewer mistakes when showing its work

Examples:

  • Instead of: "What's 15% of $4,847?"

  • Try: "What's 15% of $4,847? Let's calculate this step by step."

  • Instead of: "Help me decide between electric and gas cars"

  • Try: "I need to choose between electric and gas cars. Let's analyze this step by step: What are my driving needs? What are the costs? What about environmental impact?"

Real-World Applications

Learning and Education

Personalized tutoring: "I'm struggling with algebra and learn best through visual examples. Explain linear equations by: 1) showing the visual representation, 2) using a real-world cost calculation example, 3) walking through 3 practice problems, 4) giving me 2 problems to try."

Study aids: "Create a study guide for the American Civil War with: timeline of events, key figures and roles, causes and consequences, 10 exam questions with answers, memory tricks for dates."

Professional Work

Email communication: "Write a professional follow-up email to a potential client who hasn't responded in a week. Requirements: friendly but not pushy tone, recap our discussion, include value proposition, clear next steps, under 150 words."

Project planning: "I'm launching an e-commerce website. Break this into phases with: specific tasks, time estimates, dependencies, potential challenges, key milestones, resources needed."

Creative Content

Social media: "I run a bakery. Create: 5 Instagram post ideas showcasing products, caption templates for different post types, one-month content calendar, local hashtag strategies, user-generated content ideas."

Writing: "Write an 800-word sci-fi story: Setting: Mars colony 2150, Character: robot chef learning emotions, Conflict: annual cooking competition, Tone: humorous and optimistic, Include: dialogue and character growth."

Advanced Techniques That Work

Role-Based Prompting

Assign the AI a specific expertise role: "You are an experienced small business consultant. A restaurant client has declining sales and rising costs. Walk me through your consultation process and provide specific recommendations."

Context-Task-Format Structure

Context: Set the scene Task: State what you want Format: Specify how you want it structured

"Context: I'm interviewing for a marketing manager role at a tech startup. Task: Help me prepare thoughtful questions for the interviewer. Format: 8 questions in 4 categories (culture, role, growth, challenges) with 2 questions each and explanations of why each question is effective."

Troubleshooting Common Problems

Problem: Responses Are Too Generic

Solution: Add specificity and constraints Instead of: "Give me marketing tips" Try: "I run a pet grooming business competing with 4 others in town. Give me 5 specific, low-cost marketing tactics I can implement this month with step-by-step instructions."

Problem: Wrong Tone

Solution: Provide tone examples "Write in a professional but approachable tone—like a knowledgeable friend giving advice. Confident but not arrogant, helpful but not condescending."

Problem: Inconsistent Results

Solution: Use detailed templates "Follow this structure for every response: 1) Hook (interesting question), 2) Problem identification, 3) Solution explanation, 4) Implementation steps, 5) Expected outcomes."

Problem: Wrong Length

Solution: Be very specific "Write exactly 3 paragraphs with 4-5 sentences each. First paragraph introduces the topic, second explains the main points, third concludes with actionable advice. Target 250-300 words total."

Best Practices for Success

1. Start Crystal Clear

Before writing your prompt, define:

  • What exactly do you want as output?
  • Who is this for?
  • What tone and style do you need?
  • How long should it be?
  • What format works best?

2. Use Examples When Quality Matters

Show the AI what "good" looks like: "Write reviews like this example: 'The ErgoChair transformed my work setup. After 6 months of daily use, the lumbar support still feels perfect, and assembly took just 20 minutes. Worth every penny for desk workers. 4.5/5 stars.'"

3. Iterate and Improve

Don't expect perfection on the first try:

  1. Start simple
  2. Analyze what worked and what didn't
  3. Add more specificity where needed
  4. Test variations
  5. Save successful patterns

4. Handle Edge Cases

Anticipate problems in your prompt: "Create a 7-day meal plan for a family of 4, $100 budget, one vegetarian, one gluten-sensitive person, 30 minutes max prep time. Include grocery list by store section. If any meal exceeds time limit, provide prep-ahead modifications."

The Psychology of Effective Prompting

AI is great at:

  • Following patterns and examples
  • Generating content in established formats
  • Working with clear, specific instructions
  • Combining information in new ways

AI struggles with:

  • Highly nuanced judgment calls
  • Understanding context you didn't provide
  • Tasks requiring real-world experience
  • Implicit understanding

The Principle of Specificity: The more specific you are, the better your results. This applies to content (fitness → strength training for office workers), audience (simple → explain to a 12-year-old), format (list → numbered list with 7 items), and outcomes (help me decide → compare A vs B with 3 pros/cons each, then recommend).

Putting It All Together: A Complete Example

Poor prompt: "Help me with my presentation"

Better prompt: "I'm creating a 15-minute presentation for small business owners about social media marketing. My audience has limited marketing experience and wants practical advice they can implement immediately.

Create:

  1. Presentation outline with 5 main sections and time allocations
  2. Opening hook to grab attention in the first 30 seconds
  3. 3 specific tactics they can start this week
  4. Common mistakes section with solutions
  5. Closing call-to-action
  6. 5 potential Q&A questions with answers

Tone: Encouraging and practical, not overwhelming. Focus on simple, actionable advice over theory."

This prompt works because it provides context (audience, purpose, constraints), specific tasks, format requirements, and tone guidance.

Key Takeaways

  1. Prompting is communication: The same skills that make you good at communicating with humans apply to AI
  2. Specificity wins: Detailed prompts consistently outperform vague ones
  3. Examples are powerful: Showing works better than just describing
  4. Step-by-step thinking improves results: Ask AI to show its work for complex problems
  5. Iteration is normal: Refine your prompts until you get what you need
  6. Context matters: Provide relevant background for better responses
  7. Practice makes perfect: Start simple and gradually work up to complex applications

Remember: AI is a powerful tool, but like any tool, your skill in using it determines your results. Start with simple prompts, experiment with different approaches, and gradually work up to more sophisticated applications.

In our AI-integrated world, prompting skills are becoming as fundamental as email or internet search skills. The investment in learning these communication techniques pays dividends across virtually every area where you might use AI. From personal productivity to professional tasks to creative projects.

Sources and Further Reading

Brown, T. B., Mann, B., Ryder, N., Subbiah, M., Kaplan, J., Dhariwal, P., … Amodei, D. (2020). Language Models are Few-Shot Learners. Advances in Neural Information Processing Systems (NeurIPS 2020). https://arxiv.org/abs/2005.14165

Kojima, T., Gu, S., Reid, M., Matsuo, Y., & Iwasawa, Y. (2022). Large Language Models are Zero-Shot Reasoners. arXiv:2205.11916. https://arxiv.org/abs/2205.11916

Wei, J., Wang, X., Schuurmans, D., Bosma, M., Xia, F., Chi, E., … Zhou, D. (2022). Chain-of-Thought Prompting Elicits Reasoning in Large Language Models. arXiv:2201.11903. https://arxiv.org/abs/2201.11903

Wang, X., Wei, J., Schuurmans, D., Le, Q. V., Chi, E., & Zhou, D. (2022). Self-Consistency Improves Chain-of-Thought Reasoning in Language Models. arXiv:2203.11171. https://arxiv.org/abs/2203.11171

OpenAI. (2023). Best practices for prompt engineering with OpenAI API. OpenAI Documentation. https://platform.openai.com/docs/guides/prompt-engineering

White, J., Fu, Q., Hays, S., Sandborn, M., Gilbert, H., Olea, C., … Schmidt, D. C. (2023). A Prompt Pattern Catalog to Enhance Prompt Engineering with ChatGPT. arXiv:2302.11382. https://arxiv.org/abs/2302.11382