Skip to main content

Common Mistakes

Even experienced users can fall into common prompt engineering traps. Learn to recognize and avoid these pitfalls to get better results with less frustration.

1. Being Too Vague​

The Problem​

Generic or unclear requests lead to generic or off-target responses.

❌ Bad Example​

Tell me about this document.

✅ Good Example​

Analyze this quarterly sales report and identify:
1. Top 3 revenue drivers
2. Any concerning trends
3. Products underperforming compared to last quarter
4. 3 specific recommendations to improve Q4 results

Why It Matters​

The AI can't read your mind. Specific requests produce specific, actionable responses.

2. Missing Context​

The Problem​

Without context, the AI can't tailor its response to your situation, audience, or constraints.

❌ Bad Example​

Write a proposal for a new software system.

✅ Good Example​

Write a proposal for a new CRM software system for our 25-person Swiss financial advisory firm.

Context:
- Target audience: Board of directors (non-technical)
- Current situation: Using Excel and email, inefficient client tracking
- Budget constraint: CHF 50,000 initial + max CHF 1,000/month
- Must comply with Swiss data privacy regulations (FADP)
- Need: Better client communication tracking and automated reporting

Format: Executive summary + problem statement + proposed solution + costs + implementation timeline

Why It Matters​

Context allows the AI to make appropriate assumptions about tone, depth, priorities, and constraints.

3. Asking for Everything at Once​

The Problem​

Overwhelming prompts with multiple complex requests often lead to superficial or incomplete responses.

❌ Bad Example​

Analyze this business plan and tell me if it's good, what the market looks like, if the financial projections are realistic, what the risks are, how to improve it, and compare it to competitors, and also write a summary for investors.

✅ Good Example (Break into Steps)​

Step 1:
Review this business plan and identify the 5 most critical assumptions underlying the financial projections.

[After receiving response]

Step 2:
Now focus on the market analysis section. How does this compare to current Swiss market conditions for this industry?

[After receiving response]

Step 3:
Based on your previous analysis, create an executive summary highlighting key strengths and top 3 risks for potential investors.

Why It Matters​

Breaking complex tasks into smaller steps produces more thorough, accurate, and useful responses for each component.

4. Uploading Data Without Explanation​

The Problem​

Raw data files or documents without context leave the AI guessing about what's important and what you're looking for.

❌ Bad Example​

[Uploads Excel spreadsheet]
What do you think?

✅ Good Example​

[Uploads Excel spreadsheet]

This spreadsheet contains our Q1-Q3 sales data with the following columns:
- Date of sale
- Product category
- Units sold
- Revenue (CHF)
- Sales region (Swiss cantons)
- Sales representative

Please analyze this data and:
1. Identify which product categories are growing vs. declining
2. Highlight any seasonal patterns
3. Compare performance across regions
4. Flag any anomalies or data quality issues

Focus particularly on trends that could inform our Q4 inventory planning.

Why It Matters​

Explaining your data helps the AI understand what each column represents, what's significant, and what insights you're seeking.

5. Not Specifying Output Format​

The Problem​

Without format guidance, you might get a long paragraph when you needed bullet points, or vice versa.

❌ Bad Example​

Give me information about our top competitors.

✅ Good Example​

Provide a competitive analysis of our top 3 competitors in the Swiss wealth management space.

Format as a table with these columns:
- Competitor name
- Market share (%)
- Key differentiators
- Pricing model
- Target customer segment
- Main strengths
- Main weaknesses

After the table, provide 3 strategic recommendations based on the competitive landscape.

Why It Matters​

The right format makes information easier to use immediately, whether you need to present it, include it in a report, or make quick decisions.

6. Ignoring Token/Cost Optimization​

The Problem​

Inefficient prompts waste tokens and increase costs unnecessarily.

❌ Bad Example​

[Pastes entire 50-page report]
Summarize this document.

✅ Good Example​

Summarize the key findings from pages 15-20 of this report, which focus on Q3 financial performance.

[Paste only the relevant 5 pages]

Focus your summary on:
- Revenue changes vs. Q2
- Cost variations
- Notable risks mentioned

Keep it to 5 bullet points maximum.

Why It Matters​

Being selective about what you include and what you ask for keeps costs down while often producing better, more focused results.

7. Not Providing Examples When Style Matters​

The Problem​

Describing the style you want is less effective than showing an example.

❌ Bad Example​

Write this in a professional but friendly tone.

✅ Good Example​

Rewrite this email in our company's communication style. Here's an example of our preferred tone:

[Example email showing desired style]

Now apply that same tone and style to this new message:
[New content to be rewritten]

Why It Matters​

Examples are worth a thousand words of description when it comes to style, tone, and format.

8. Forgetting the Audience​

The Problem​

Not specifying who will read or use the output leads to inappropriate language, detail level, or focus.

❌ Bad Example​

Explain our Q3 results.

✅ Good Example​

Explain our Q3 results for a board presentation.

Audience: Non-technical board members with financial literacy
Focus: Strategic implications rather than operational details
Tone: Professional, data-driven, balanced (acknowledge both successes and challenges)
Length: 3 minutes of speaking time (approximately 400 words)

Why It Matters​

The same information needs to be communicated very differently to the board vs. the technical team vs. external stakeholders.

9. Not Iterating Based on Results​

The Problem​

Treating each prompt as a one-shot attempt instead of a conversation.

❌ Bad Approach​

[Gets unsatisfactory response]
"This AI doesn't understand what I need."
[Gives up or starts completely over]

✅ Good Approach​

[Gets initial response]
"This is helpful, but too technical. Can you rewrite it for a non-technical audience, and focus more on business implications than technical details?"

[Gets improved response]
"Better! Now add 3 specific action items we can implement this quarter."

Why It Matters​

AI conversations are iterative. Each exchange helps you refine toward exactly what you need.

10. Overlooking Validation​

The Problem​

Accepting AI-generated content without review, especially for high-stakes or specialized topics.

⚠️ Remember​

Always validate AI outputs, especially for:
- Legal or regulatory content
- Financial calculations and projections
- Technical specifications
- Medical or safety-critical information
- Client-facing materials

Best Practice​

Use AI as a powerful first draft and analysis tool, but always apply human judgment and expertise for final review, particularly in:
- Specialized domains requiring expert knowledge
- Contexts with legal or compliance implications
- Situations where errors could have significant consequences

Why It Matters​

AI is a tool to augment human expertise, not replace it. Critical thinking and domain expertise remain essential.

Quick Reference: Avoiding Common Mistakes​

MistakeQuick Fix
Too vagueAdd specific deliverables and success criteria
Missing contextInclude who, what, when, where, why
Too complexBreak into multiple smaller prompts
Unexplained dataDescribe what the data represents and what you're looking for
Wrong formatSpecify exact structure (bullets, paragraphs, table, etc.)
Token wastePaste only relevant portions; ask focused questions
Unclear styleProvide an example of the desired style
Wrong audienceSpecify who will use/read the output
No iterationTreat prompts as conversations, refine as needed
No validationAlways review and verify critical outputs

Practice Exercise​

Try rewriting this poor prompt using what you've learned:

❌ Bad Prompt:

Look at this data and tell me what you think.

✅ Improved Version (Your Turn):

Think about:

  • What specific insights do you want?
  • What context does the AI need?
  • What format would be most useful?
  • Who is the analysis for?
  • What should the AI focus on or avoid?

Mastering these common mistakes will dramatically improve your AI interactions and results.