Prompt Engineering 101: Getting Better Results from ChatGPT and Claude

Why Your AI Results Are Disappointing

You've tried ChatGPT. Maybe Claude. The results were... fine. Sometimes useful, often generic, occasionally completely wrong.

Here's the thing: the AI isn't the problem. Your prompts are.

Prompt engineering is the skill of communicating with AI systems effectively. It's the difference between getting generic fluff and getting outputs that genuinely save you time.

Let's fix your prompts.

The Five Principles of Effective Prompts

Principle 1: Be Specific About What You Want

Bad prompt: "Write me a marketing email"

Good prompt: "Write a marketing email for our B2B SaaS product (project management for construction companies). The email should announce our new mobile app feature, target project managers at mid-size construction firms, be 150-200 words, and include a clear CTA to book a demo."

The good prompt gives the AI everything it needs to produce relevant output on the first try.

Principle 2: Provide Context

AI doesn't know your business, your audience, or your constraints. You need to tell it.

Template:

Context: [Describe your business/situation]
Audience: [Who will read/use this]
Constraints: [Word count, tone, format requirements]
Goal: [What you're trying to achieve]

Principle 3: Use Examples (Few-Shot Prompting)

Show the AI what good output looks like.

Example: "Write product descriptions in this style:

Example 1: Product: Wireless earbuds Description: 'Immersive sound meets all-day comfort. Our wireless earbuds deliver studio-quality audio with 8 hours of battery life, so your music never stops.'

Example 2: Product: Running shoes
Description: 'Engineered for the long run. Responsive cushioning and breathable mesh keep you moving mile after mile.'

Now write a description for: Laptop backpack"

Principle 4: Assign a Role

Tell the AI who it should be.

Example: "You are a senior financial analyst at a Fortune 500 company. Analyze this quarterly report and identify the three most significant trends."

Role assignment primes the AI to use relevant knowledge and appropriate tone.

Principle 5: Break Complex Tasks into Steps

For complex outputs, use chain-of-thought prompting.

Example: "I need to create a pricing strategy for our new product. Let's work through this step by step:

Step 1: First, analyze these three competitor prices and identify where we could position: [data]

Step 2: Based on our costs of $X, calculate what margins we'd achieve at different price points.

Step 3: Recommend a price point with justification."

Advanced Techniques

The Persona Stack

Layer multiple personas for more nuanced output:

"First, analyze this business plan as a skeptical venture capitalist looking for weaknesses. Then, analyze it as an optimistic entrepreneur looking for opportunities. Finally, synthesize both perspectives into balanced feedback."

Iterative Refinement

Don't accept the first output. Use follow-up prompts:

• "Make this more concise"
• "Add more specific examples"
• "Adjust the tone to be more conversational"
• "What are three ways this could be improved?"

Output Formatting

Specify exactly how you want the output structured:

"Format your response as:

1. Executive Summary (2-3 sentences)
2. Key Findings (bullet points)
3. Recommendations (numbered list)
4. Next Steps (action items with owners)"

Common Mistakes to Avoid

Mistake 1: Being Too Vague

Generic prompts get generic responses. Always include specifics.

Mistake 2: Asking for Too Much at Once

Break large requests into smaller, focused prompts.

Mistake 3: Not Iterating

Your first prompt rarely produces perfect output. Plan to refine.

Mistake 4: Ignoring Context Windows

AI has limited memory. For long conversations, periodically summarize the key points.

Mistake 5: Trusting Without Verifying

AI can be confidently wrong. Always fact-check important outputs.

Prompts for Business Use Cases

Meeting Notes → Action Items

"Review these meeting notes and extract: 1) Decisions made, 2) Action items with owners and due dates, 3) Open questions requiring follow-up. Format as a table."

Email Response

"Draft a response to this email that: acknowledges their concern, provides a solution, maintains a professional but warm tone, and is under 100 words. Email: [paste email]"

Data Analysis

"Analyze this sales data and identify: 1) Top 3 trends, 2) Any anomalies that warrant investigation, 3) Predictions for next quarter based on patterns. Explain your reasoning."

Content Repurposing

"Transform this blog post into: 1) A LinkedIn post (under 300 words), 2) Three tweet-length insights, 3) An email newsletter summary. Maintain the key message but adapt tone for each platform."

The Meta-Skill

The best prompt engineers constantly experiment. They:

1. Save prompts that work well
2. Iterate on prompts that don't
3. Build prompt libraries for common tasks
4. Stay updated on new model capabilities

Prompt engineering is a skill that compounds. Every hour invested pays dividends across thousands of future interactions.

---

Want to level up your team's prompt engineering skills? Our training programs turn novices into power users in days, not months.