Prompt Engineering Lab — Learn to Talk to AI

Prompt Engineering Lab — Learn to Talk to AI
Ariel Levin Founder @ SDR Labs, Generative AI Enthusiast SDRLabs.io
The Agency Advantage Conference June 25, 2025 | 11 AM ET
AL
by Ariel Levin
 
About Ariel Levin
Agency Owner
Running SDR Labs with hands-on experience in digital marketing
Daily AI User
Using generative AI for:
Content creation
Account and lead research
Lead qualification and scoring
Proven Track Record
My LinkedIn post about prompts reached 5.2M views
Heavy Practitioner
Used millions of tokens across various AI platforms
Real-World Experience
What I share today comes from actual agency operations, not theory
Agency AI Use Cases
Why Prompt Engineering Matters
Most people treat AI like Google
Asking casual questions and getting mediocre, generic responses
Inconsistent quality without structure
Different results every time makes AI unreliable for business use
AI doesn't know your context
It doesn't understand your industry, company, style, or specific needs
Wasted time and money
Using expensive models for simple tasks, or getting poor results that need redoing
Can't scale across teams
Without proper frameworks, you can't systematize AI workflows for multiple people
Amateur vs. professional results
The difference between "write me an email" and engineered prompts with clear objectives
What You'll Learn Today
Master the three types of prompts
Non-reasoning, reasoning, and search prompts, plus when to use each
Professional 4-step creation process
My proven method for building high-quality prompts systematically
Real agency implementation examples
Lead qualification workflows and content repurposing that you can use immediately
AI Models We'll Use & The Three Types of Prompts
AI Models We'll Use
Perplexity.ai
Always use this for generating professional prompt guidelines (the foundation of our process)
(Use Pro version) 5 free runs a day
Main AI Choice
Pick one: ChatGPT, Claude, or Gemini for actually running your prompts and daily work.
My recommendation: Use ChatGPT O3, It follows guidelines more dilligently
Why these tools
Perplexity gives the best guidelines, while the big three offer reliable performance for agency workflows
The Three Types of Prompts
Non-Reasoning Prompts
Standard prompts for most daily tasks like content creation and basic outreach
Reasoning Prompts
Advanced prompts that make AI "think" step-by-step for complex analysis and decisions
Search Prompts
Prompts designed to find and retrieve specific information about prospects or topics
Understanding Each Prompt Type
Non-Reasoning Prompts
When to use them: Daily tasks like content creation, email writing, social media posts, Analyzing client research and basic outreach
Why they're your go-to choice: Fast execution, cost-effective, and high quality for standard agency work
What makes them effective: Direct input-to-output processing without complex analysis or "thinking" steps
Reasoning Prompts
When you need a coach or mentor: Getting strategic advice, problem-solving guidance, or expert recommendations
When you need special resources or plans: Creating customized strategies, detailed campaign plans, or complex analyses
When you want thought leader content: Developing insightful perspectives, industry analyses, or expert-level commentary
And In B2B when you want to give smart relevant personalized advice
Search Prompts
Prospect research: Check websites, PR mentions, testimonials, local coverage, and company background
Up-to-date information: Get current market trends, recent news, pricing, or industry developments
Thorough research tasks: Deep-dive analysis requiring multiple sources and comprehensive data gathering
The Professional Process & AI Projects
The Professional Process
4-Step Method
Guidelines → Creation → Improvement → Implementation
Professional-grade results
This systematic approach creates reliable, high-quality prompts vs amateur attempts
Live Demo
I'll show you this process in action with a real agency use case
Agency Use Cases by Prompt Type
Non-Reasoning Example - Convert YouTube transcript into LinkedIn post, Facebook post, Twitter thread, and Instagram caption
Reasoning Example - Analyze prospect research data to determine if they match your ICP and recommend best approach strategy
Search Example - Research prospect's website, recent PR, testimonials, and local mentions to gather background for outreach
AI Projects - Beyond Single Prompts
When to use Projects
For ongoing work with clients, specialized workflows, or when you need AI to remember your context
The problem they solve
Stop re-explaining who you are, your company, and your style in every conversation
Context window limits
Long conversations make AI "forget" earlier information, Projects maintain consistency
Project Setup Components
Knowledge Base
Upload your resume, company info, work examples, and client-specific guidelines
Guidelines (Uploaded to the KB)
Specific instructions for different types of content (cold outreach, responses, industry-specific styles).
Guidelines could even be code or a formula (for example to calculate ROI)
Instructions
Overall directions including purpose, what's in knowledge base, when to use code and other tools, and workflow process
The 4 Professional Prompts
Step 1: Guidelines Generation (In perplexity)
Write the guidelines for a prompt engineer AI agent who writes [TYPE] prompts.
(Could be claude or chatgpt)
Where [TYPE] is: non reasoning prompts, reasoning prompts, or search prompts
Step 2: Prompt Creation Using Guidelines (In Any LLM)
Write the following prompt following these [guidelines:
[YOUR SPECIFIC TASK DESCRIPTION]
[PASTE GUIDELINES HERE]
Step 3: Prompt Improvement Request
Suggest ways for me to improve this prompt.
Step 4: Implementation of Improvements
Rewrite the prompt implementing all suggestions.
Rename the chat to easily located it 
Key Takeaways & Next Steps
Key Takeaways
Projects are for lazy people
Set them up once with your company info, examples, and guidelines to save massive time
Use the 4-step professional process
Guidelines → Creation → Improvement → Implementation creates reliable, high-quality results
Match prompt type to task
Non-reasoning for daily work, reasoning for strategy, search for research
Implementation Next Steps
Start with one repetitive task
Pick email outreach, content creation, or prospect research that you do daily
Use the 4-step process
Go to Perplexity, generate guidelines, create your first professional prompt, then improve it
Set up your first Project
Upload your company info and examples so AI remembers your context and style
Thank You
Ariel Levin
LinkedIn: ariellevin 
Email: [email protected] 
Website: www.sdrlabs.io
Made with