You’ve experimented with AI prompts. Sometimes they deliver precisely what you need. Sometimes they miss the mark entirely. The difference isn’t the AI tool—it’s how you communicate with it.
Most people write AI prompts the way they’d search Google: a few keywords and hope for the best. That’s why they get inconsistent results and conclude that AI tools are overhyped or unreliable.
This guide teaches 15 specific techniques that consistently produce useful outputs from AI prompts. Each includes before-and-after examples that show exactly what changes and why they work, helping you transform vague requests into precise instructions that deliver results.
Table of Contents
Why Most Prompts Fail
Before learning what works, understand why prompts fail:
Failure Pattern 1: Too Vague “Write an email” → ChatGPT has no idea what email, to whom, about what, or what tone
Failure Pattern 2: Missing Context “Draft proposal” → ChatGPT doesn’t know your business, the client, the project, or the requirements
Failure Pattern 3: Wrong Tool for Task “Calculate my profit margin” → ChatGPT is terrible at maths, brilliant at explaining concepts
Failure Pattern 4: No Refinement Accept first mediocre output → Miss the 2-3 iterations that create quality
Failure Pattern 5: Unrealistic Expectations Expect perfection from vague instructions → Get frustrated, quit
The pattern: Poor inputs create poor outputs. Great prompts create great results.
Technique 1: The Five-Element Structure
What it is: Every prompt answers five questions
The elements:
- What do you want? (email, summary, list, analysis)
- Who is it for? (audience, reader, user)
- How long? (word count, paragraphs, time)
- What tone? (professional, casual, formal, friendly)
- Specific requirements? (points to include, things to avoid)
Bad prompt:
Write a social media post
Good prompt:
Write a 150-word LinkedIn post for Belfast small business owners about time management—professional but conversational tone. Include one practical tip they can implement today and end with a question to drive engagement.
Why it works: Eliminates ambiguity. ChatGPT knows precisely what you need.
Common failure fix: “I got generic output” → You gave generic instructions. Add the five elements.
Technique 2: Context First
What it is: Start conversations with background information
Why it matters: ChatGPT doesn’t know you, your business, or your industry. Every conversation starts fresh.
Bad approach:
[First prompt]: Write an email to the customer
[Second prompt]: No, that’s too formal
[Third prompt]: Make it more casual
[Fourth prompt]: Add information about our refund policy
Wastes time re-explaining with each prompt.
Good approach:
Context: I run a small online retail business in Belfast, selling sustainable homeware. Our customers are environmentally-conscious, aged 25-45. Our tone is friendly and approachable, not corporate. We have a 30-day money-back guarantee, no questions asked.
Now, draft a 150-word email to the customer asking about the return process…
Why it works: One context-setting prompt improves every subsequent prompt in that conversation.
Before/After Example:
Without context: Generic corporate email about “processing your return request”
With context: Warm, personal email matching your brand: “We’re sorry the ceramic bowl wasn’t quite right for you! Here’s how our hassle-free 30-day return works…”
Time saved: 10-15 minutes of back-and-forth clarification
Technique 3: Show, Don’t Just Tell
What it is: Provide examples of what you want
Why it matters: “Professional tone” means different things to different people. An example shows precisely what you mean.
Bad prompt:
Write in my brand voice
Good prompt:
Write in my brand voice. Here’s an example of our style:
“Coffee that doesn’t require a degree to order. Just great beans, properly made, served by people who remember your name. Belfast’s friendliest caffeine dispensary.”
Match this conversational, slightly irreverent tone and sentence structure.
Why it works: ChatGPT matches patterns. Give it the pattern you want.
What you can show:
- Style examples: Paste your previous writing
- Format examples: Show the structure you need
- Tone examples: Demonstrate voice
- Similar content: “Like this but about [different topic]”
Common failure fix: “It doesn’t sound like me” → Show it what “like you” sounds like
Technique 4: Specify Constraints
What it is: Limitations that force better outputs
Why it works: Constraints create focus and eliminate fluff
Powerful constraints:
Word-level:
- “Using only words a 12-year-old would understand”
- “No jargon or technical terms”
- “Avoid these overused phrases: [list]”
Structural:
- “Maximum 3 sentences per paragraph”
- “Each bullet point 15-25 words”
- “No paragraph over 50 words”
Stylistic:
- “Without using: innovative, cutting-edge, revolutionary, leverage, synergy”
- “As if explaining to your grandmother”
- “No corporate speak whatsoever”
Before:
Write a product description for project management software
Generic output full of buzzwords: “Leverage cutting-edge technology to revolutionise your workflow…”
After:
Write a 150-word product description for project management software. Use only words a 14-year-old would understand. No business jargon. Each sentence is under 15 words. Explain what it does, why it helps, and what makes it different. Avoid these words: innovative, cutting-edge, revolutionary, synergy, and leverage.
Clear, accessible description: “See all your projects in one place. Know who’s working on what. Never miss a deadline. Simple enough to start using in 5 minutes…”
Why it works: Forces clarity and simplicity
Technique 5: Request Multiple Options
What it is: Ask for variations instead of one answer
Why it works: Compare approaches, choose the best elements, or combine ideas
Bad prompt:
Write email subject line
Good prompt:
Write 5 email subject line variations:
1. Direct
2. Question-based
3. Benefit-focused
4. Urgency/scarcity angle
5. Curiosity-driven
Each 6-8 words maximum. I’ll test the most promising.
Why it works:
- Tests different approaches simultaneously
- Identifies what resonates
- Combines the best elements
Applications:
Content angles: “Give me 5 different hooks for this article: emotional, logical, story-based, question-based, contrarian”
Messaging: “Create 3 versions emphasising: cost savings, time savings, quality improvement”
Formats: “Present this information as: bullet points, narrative, Q&A, step-by-step, comparison table”
Time saved: Testing 5 approaches in 2 minutes instead of writing each individually for over an hour
Technique 6: Structure the Output
What it is: Tell ChatGPT exactly how to format the response
Why it works: Eliminates manual reformatting
Bad prompt:
Analyse this customer feedback
Unstructured wall of text you need to organise.
Good prompt:
Analyse this customer feedback [paste]. Structure response as:
1. Overall sentiment summary (2 sentences)
2. Table with columns: Theme | Frequency | Example Quote
3. Top 3 positive themes
4. Top 3 concerns
5. Recommended actions (numbered list)
6. Priority order with justification
Ready-to-use formatted analysis.
Why it works: Information is organised how you need it from the start
Format specifications:
Tables: “Present as a table with 4 columns: [names]. Include header row.”
Hierarchical: “Format as: Main header (H2), 3-5 sections (H3), bullet points under each, conclusion”
Lists: “Numbered list with bold subheadings and 2-3 sentence explanations”
Comparative: “Side-by-side comparison: Option A | Option B with rows for: Cost, Time, Pros, Cons, Recommendation”
Common failure fix: “The information is good, but formatting is messy” → Specify format in the prompt
Technique 7: Iterative Refinement
What it is: Never accept the first output as final. Refine 2-3 times.
Why it matters: First outputs are 70-80% there. Iteration reaches 95%.
The process:
Prompt 1 (Initial):
Write a 300-word blog introduction about ChatGPT for small businesses
Prompt 2 (Refine):
Good start. Make it more conversational and less formal. Replace the first paragraph with a hook that addresses a specific pain point.
Prompt 3 (Polish):
Better. Now, reduce to 200 words, maintaining the conversational tone while making every sentence more concise and impactful. Remove any filler.
Why it works: Each iteration improves specific aspects without starting over
Common refinement prompts:
Length:
- “Reduce to 150 words maximum”
- “Expand with more examples”
- “Cut 30% while keeping key points”
Tone:
- “Make more casual/formal”
- “More enthusiastic/measured”
- “Less sales-y, more educational”
Focus:
- “Focus specifically on [narrow aspect]”
- “Emphasise benefits over features”
- “Lead with the biggest problem this solves”
Structure:
- “Rewrite as bullet points”
- “Break into shorter paragraphs”
- “Add subheadings”
Common failure fix: “The output is close but not quite right” → That’s normal. Refine it instead of starting over.
Technique 8: Role Assignment
What it is: Tell ChatGPT what expertise or perspective to adopt
Why it works: Different roles produce different insights and styles
Bad prompt:
Give me marketing advice
Good prompt:
Act as an experienced marketing consultant specialising in Belfast small businesses with limited budgets. You’ve helped over 50 local companies grow through practical, cost-effective strategies.
Given this context, advise me on…
Why it works: The role shapes how ChatGPT analyses and responds
Effective roles:
Expert roles:
- “You’re a seasoned accountant who explains complex concepts simply”
- “Act as a digital marketing specialist with 15 years of experience”
- “Respond as an HR director familiar with small business constraints”
Customer roles:
- “Take the perspective of a sceptical potential customer”
- “Act as someone who’s never heard of this product”
- “Think like a budget-conscious small business owner”
Adversarial roles:
- “You’re a competitor analysing my strategy. What weaknesses do you see?”
- “Act as a critical reviewer. What’s wrong with this approach?”
- “You’re a customer who had a bad experience. What frustrates you?”
Before/After Example:
No role: “Marketing advice for new business” → Generic marketing platitudes
With role: “Act as a Belfast business advisor who’s launched 30+ local businesses. Marketing advice for a new café in the University Quarter.” → Specific, locally-relevant, practical advice about the student market, local competition, and seasonal patterns
Common failure fix: “Advice is too generic” → Assign a specific expert role
Technique 9: Negative Instructions
What it is: Tell ChatGPT what NOT to do
Why it works: Prevents common unwanted patterns
Bad prompt:
Write product benefits
Gets a feature dump full of jargon.
Good prompt:
Write 5 key product benefits.
Do NOT:
– List features (focus on outcomes)
– Use jargon (plain language only)
– Make unverifiable claims (“best,” “revolutionary”)
– Write in corporate speak
– Exceed 25 words per benefit
Clean, benefit-focused copy.
What to prohibit:
Style elements: “Avoid: emojis, exclamation marks, ALL CAPS, corporate buzzwords”
Overused phrases: “Don’t use: leverage, synergy, cutting-edge, innovative, game-changing, paradigm shift”
Structural issues: “No paragraphs over 50 words, no sentences over 20 words, no passive voice”
Content elements: “Don’t include: pricing, competitor names, unverified statistics, technical specifications”
Why it works: Steers away from ChatGPT’s tendency toward specific patterns
Technique 10: Chain Prompting
What it is: Break complex tasks into sequential steps
Why it works: Focused requests produce better results than complex multi-part requests
Bad approach (one massive prompt):
Create a complete content marketing strategy, including: audience analysis, content calendar, 10 blog topics, 3 full blog posts, social media plan, email sequence, and metrics to track
Rushed, superficial, everything.
Good approach (chain of focused prompts):
Prompt 1:
Analyse the target audience for [product]. Include: demographics, pain points, where they seek information, and buying triggers.
Prompt 2:
Based on that audience analysis, create a 3-month content calendar. Include: weekly themes, content types, channels, calls-to-action.
Prompt 3:
From that calendar, develop a detailed outline for Month 1, Week 1 blog post about [topic].
Prompt 4:
Write a complete blog post from that outline. 1,500 words, [specifications].
Prompt 5:
Create 5 social media posts promoting that blog post. [Specifications for each platform].
Each builds on the previous, maintains consistency, and achieves quality.
Why it works:
- ChatGPT maintains context within conversation
- Each step is focused and thorough
- You verify quality before building on it
- Easy to adjust mid-process
Common failure fix: “Trying to do too much in one prompt” → Break into 3-5 sequential steps
Technique 11: Provide Data, Request Insight
What it is: Give ChatGPT information, ask for an interpretation
Why it works: ChatGPT excels at analysis but struggles with facts. Provide facts, get analysis.
Bad prompt:
What should I do about declining sales?
Generic advice without understanding your situation.
Good prompt:
Here’s my sales data for 6 months:
Jan: £15,200
Feb: £14,800
Mar: £16,100
Apr: £13,900
May: £12,200
Jun: £11,800
Context: Retail business, no major changes to operations or competition.
Analyze:
1. What trends do you see?
2. What typically causes this pattern?
3. What 3 questions should I investigate?
4. What actions might reverse the trend?
Specific, relevant analysis of YOUR data.
Why it works: You provide facts (accurate), ChatGPT provides interpretation (valuable)
Applications:
Customer feedback: [Paste 20 customer comments] “Identify common themes, sentiment, and priorities for improvement”
Financial data: [Paste monthly figures] “Analyse trends, identify anomalies, suggest areas for investigation”
Competitor information: [Paste notes from competitor analysis] “Compare approaches, identify their strategy, suggest differentiation opportunities”
Survey results: [Paste survey data] “Summarise findings, segment by customer type, recommend actions”
Common failure fix: “ChatGPT gave wrong numbers” → Don’t ask it to calculate. Provide it with numbers and ask for an interpretation.
Technique 12: Critique and Improve
What it is: Have ChatGPT critique its own output, then rewrite
Why it works: Self-analysis identifies weaknesses you might miss
The process:
Step 1: Get initial output
Step 2:
Critique what you just wrote:
– Is the tone appropriate for the audience?
– Are there vague or generic statements?
– Could any section be more specific or actionable?
– What’s missing that would add value?
– Rate overall quality 1-10 and explain the rating
Step 3:
Now rewrite it, addressing all those weaknesses
Why it works: Version 2 is consistently 30-40% better than Version 1
Before/After Example:
Version 1 (initial): “Our product helps businesses save time and increase productivity through innovative solutions.”
After self-critique: “Too vague. ‘Save time’ how? ‘Productivity’ in what way? ‘Innovative solutions’ is a meaningless buzzword. No specific benefits. Rating: 4/10”
Version 2 (rewrite): “Our software automates invoice processing, reducing 4 hours of weekly data entry to 30 minutes. Belfast businesses using it process 85% more invoices with the same team.”
Specific, quantified, meaningful.
Technique 13: Format with Examples
What it is: Show the exact format you want
Why it works: Eliminates format ambiguity
Bad prompt:
Create an FAQ about our service
Good prompt:
Create an FAQ about our service. Format exactly like this:
**Q: [Question in customer’s words, 10-15 words]**
A: [Concise answer, 50-75 words, addressing concern directly and ending with clear next step]
Create 5 FAQs following this exact format covering: pricing, implementation time, support, training, and cancellation policy.
Perfect formatting from the start.
What to show examples of:
Headers and structure: “Format like: ## Main Heading, ### Subheading, bullet points under each”
Lists and bullets: “• Bullet format: Bold Headline (20 words max) followed by 2-3 sentence explanation”
Q&A pairs: Show the exact question and answer structure you want
Tables: “| Column 1 | Column 2 | Column 3 |” with example row
Why it works: ChatGPT is excellent at pattern matching. Show the pattern.
Technique 14: Audience Perspective Shift
What it is: Request the same content from different audience viewpoints
Why it works: Reveals different value propositions and messaging angles
The prompt:
Explain [product/service] from three different perspectives:
1. Technical user (focus: features, capabilities, integration)
2. Business owner (focus: ROI, time savings, competitive advantage)
3. End user (focus: ease of use, immediate benefits, daily impact)
150 words each, no overlap in key points.
Why it works:
- Identifies which benefits matter to which audiences
- Creates segmented marketing messages
- Ensures you’re speaking to everyone’s concerns
Applications:
Marketing pages: Create different sections for different visitor types
Sales materials: Prepare for different stakeholder concerns
Training content: Address what each role cares about
Product positioning: Understand multi-audience appeal
Technique 15: Specify the “So What?”
What it is: Force ChatGPT to focus on implications and outcomes, not just information
Why it works: Prevents generic information dumps
Bad prompt:
Write about email marketing
Generic overview of what email marketing is.
Good prompt:
Write about email marketing for Belfast small businesses.
For each point, answer “So what? Why does this matter?”
Focus on: specific actions they should take this week, expected results, common mistakes that cost money, and how to measure success.
Not: general definitions or theory.
Actionable, outcome-focused content.
Why it works: “So what?” forces practical application over information
Before/After:
Without “So what?”: “Email marketing involves sending promotional messages to a list of subscribers…”
With “So what?”: “Send emails Tuesday-Thursday 10 am. Tests across 1,000+ Belfast businesses show 34% higher open rates than on Mondays and Fridays. That’s 340 more people seeing your message per 1,000 subscribers. Takes 2 minutes to reschedule your campaigns…”
Information → Action → Outcome
Putting It All Together: Master Prompt Example
Here’s how multiple techniques combine:
[Context – Technique 2]
I run a 5-person marketing agency in Belfast serving local SMEs. We’re writing a proposal for a traditional family restaurant (25 years established, with no current online presence).
[Role – Technique 8]
Act as a marketing consultant who respects traditional businesses while explaining modern needs.
[What + Audience + Length – Technique 1]
Write a proposal outline (not a full proposal) for a social media management service. 500 words. Audience: restaurant owners sceptical about social media.
[Structure – Technique 6]
Format as:
1. Understanding their situation (2-3 sentences showing we get them)
2. Why social media matters for restaurants (3 specific local benefits)
3. Our 6-month approach (timeline with clear milestones)
4. What success looks like (specific, measurable outcomes)
5. Investment and next steps (transparent, low-pressure)
[Constraints – Technique 4]
Avoid: corporate jargon, over-promising, aggressive sales language
Use: respect for traditional values, local Belfast examples, practical outcomes
[Show don’t tell – Technique 3]
Tone like this: “Your restaurant has thrived for 25 years through word-of-mouth and excellent food. That foundation doesn’t change. Social media simply helps more people discover what regulars already know.”
Result: A perfect foundation for your proposal, requiring only client-specific details.
Common Failures and Fixes
“Output is generic” → Add specific context, examples, and constraints
“Wrong tone” → Show example of desired tone, specify what to avoid
“Too long/short” → Specify exact word count or structural limits
“Missing key information” → List specific points that must be included
“Doesn’t match my style” → Provide a sample of your writing to match
“Lacks structure” → Specify exact format with section headings
“Too vague” → Request specific examples, numbers, actions
“Trying to do too much” → Break into chain of focused prompts
Measuring Technique Effectiveness
Track these metrics:
First-try success rate:
- Target: 60%+ usable without iteration
- Measure: Percentage of prompts producing sound first output
Iteration count:
- Target: 2-3 refinements to publication quality
- Measure: Average iterations per final output
Time to completion:
- Target: 70-80% faster than manual
- Measure: Total time including refinement
Edit time:
- Target: 80%+ usable with minor edits only
- Measure: Time spent editing vs rewriting
Master These Techniques: AI Prompts
These 15 techniques work. But knowing them isn’t the same as using them effectively.
Our free ChatGPT Masterclass teaches you to apply these techniques:
- Live demonstrations of each technique
- Practice exercises with feedback
- Common mistakes and quick fixes
- Industry-specific applications
- 25+ ready-to-use prompts incorporating multiple techniques
The difference between frustrating ChatGPT use and transformative results is writing prompts that actually work.
These 15 techniques are used by businesses, saving over 15 hours weekly and achieving 91% revenue increases from AI adoption.
Learn them. Apply them. Refine them. That’s the path to results.
About Future Business Academy
We’re Belfast’s AI training specialists, teaching businesses across Northern Ireland and Ireland to write prompts that consistently deliver results. We focus on practical techniques rather than theoretical concepts.
For comprehensive AI implementation, our parent company, ProfileTree, provides strategic consulting and hands-on support.
