Lab: The Prompting Challenge

This will take 15-20 minutes. Set a timer if that helps. This is the most important hands-on exercise of the course so far, because it takes everything you've learned so far this week and makes it real.

You're going to do the same task three times, three different ways. By the end, the difference in quality will be so obvious you'll never go back to the lazy approach.

Let's go.

Step 1: Pick your task

Choose something meaty from your actual work. Not "write me a haiku." Something that would normally take you 30-60 minutes. Here are good options:

• Write a strategy memo or recommendation
• Analyze a competitor and summarize their strengths and weaknesses
• Create a presentation outline for an upcoming meeting
• Draft a proposal for a client or internal stakeholder
• Build a 90-day plan for a project or initiative

Pick one. Use something real. The exercise is 10x more valuable with real stakes.

Step 2: Attempt 1 — The way most people do it

Open ChatGPT or Claude. Either works.

Type your request the way you normally would. The way you'd ask if you were in a rush and just wanted something fast. Don't overthink it. Most people type something like:

"Write a strategy memo about expanding into the European market."

"Create a presentation outline for our Q1 review."

"Analyze our top 3 competitors."

Hit enter. Save the output somewhere (copy it to a doc or just keep the tab open). We'll compare later.

Step 3: Attempt 2 — Role + Context + Task + Format

Now open a new conversation. Same task. But this time, use the framework from Day 7:

Role: Start with "You are a [specific expert]." Be as specific as possible. Not "a consultant" but "a senior strategy consultant who specializes in B2B SaaS expansion into European markets."

Context: Give it 3-5 sentences of background. What's the situation? What's the company? What's been tried? What's the goal?

Task: Be specific about what you want. Not "write a memo" but "write a 1-page strategy memo recommending whether we should expand into the EU market in Q3, with supporting evidence."

Format: Tell it exactly how the output should look. "Use headers. Start with the recommendation, then the reasoning, then the risks. Keep it under 500 words."

Hit enter. Save this output too.

Step 4: Attempt 3 — Add chain of thought and stages

One more new conversation. Same task. This time, go all in.

Start with the role and context from Attempt 2. But now add:

"Think step by step."

And break it into stages:

"First, identify the 5 most important factors we should consider for this decision. Then evaluate each factor given our specific context. Then give me your recommendation. Finally, list the top 3 risks and how we'd mitigate each one."

This forces the AI to show its work, build on each piece, and give you something that has genuine depth.

Hit enter. Save it.

Step 5: Compare

Pull up all three outputs side by side. Look at them honestly.

Questions to ask yourself:

• Which one could you actually send to your boss or a client?
• Which one demonstrates real understanding of your specific situation?
• Which one would take the least editing?
• Where did the AI miss the mark in Attempt 1 that it nailed in Attempt 3?

For most people, the difference between Attempt 1 and Attempt 3 is night and day. Same AI. Same task. Different input.

Bonus: Set up persistent context

If you have a few extra minutes, do this. It pays dividends on every future conversation.

For ChatGPT: Go to Settings > Personalization > Custom Instructions. Fill in who you are, what you do, and how you like responses. Here's a template:

"I'm [role] at [company type]. I work on [what you do]. I prefer direct, concise responses. Use bullet points for lists. No filler. When I ask for analysis, think step by step."

Link: ChatGPT Custom Instructions

For Claude: Create a new Project at claude.ai. Add your context to the Project Instructions. You can also upload documents (your company overview, brand guidelines, strategy docs) that Claude will reference in every conversation within that project.

Link: Claude Projects guide

Reflection

Take 2 minutes and think about these:

1. How much better was Attempt 3 compared to Attempt 1? (If the answer is "not much," your task might have been too simple. Try again with something harder.)

1. What specific part of the framework made the biggest difference for your task? The role? The context? The step-by-step structure?

1. Going forward, what's the minimum you'll include in every prompt? (Most people land on: role + context at a minimum. It takes 15 extra seconds and saves 15 minutes of editing.)

This week you went from "I've tried ChatGPT" to understanding how the best AI users actually work. The techniques are simple. Role. Context. Task. Format. Chain of thought. Persistent instructions.

None of this is complicated. The hard part is making it a habit. Next week, we'll put these skills to work on the specific tasks you do every day.

The difference between Attempt 1 and Attempt 3 is the difference between dabbling with AI and actually being good at it.