Master AI Prompting With Role-Based Techniques and Iterative Refinement

# I AM GPTED: EPISODE SCRIPT

---

**[INTRO MUSIC: Upbeat, slightly quirky tech vibes]**

**MAL:** Hey there, I'm Mal—The Misfit Master of AI—and welcome back to *I am GPTed*, the show where we make AI actually *useful* instead of just impressive at writing bad poetry about your cat.

Look, I've spent way too much time talking to robots, and somehow I've figured out what actually works. So buckle up. Today we're covering the stuff that'll actually change how you use AI.

**[MUSIC FADES]**

**MAL:** Let's start with something called **role prompting**. Yeah, I know it sounds like corporate nonsense, but hear me out—it actually *works*.

See, most people just ask their AI something like, "Give me a job description." Fine. You get a job description. It's fine. It's boring. Everyone hates it.

Here's the before-and-after:

**Bad prompt:** "Write a job ad for a marketing manager."

**Good prompt:** "You're a ruthless startup hiring manager with a budget the size of a small country. Write a job ad that'll make talented people actually *want* to apply instead of delete the email."

Same AI. Completely different output. The second one has personality. It's punchy. It sounds like an actual human wrote it.

Why? Because you told the AI what *perspective* to take. It's like asking a chef their opinion on your ingredients versus just handing them a shopping list. The perspective changes everything.

**[TRANSITION SOUND]**

**MAL:** Now, here's something people never think about: **AI as your research buddy for non-fiction**.

You just finished a chapter on how people make terrible financial decisions. Instead of just moving on, *talk to the AI about it*. Ask it questions. Challenge it. "Wait, doesn't that contradict what neuroscience says about risk?"

Suddenly you've got this ongoing conversation where the AI is helping you spot connections you'd miss alone. It's like having a study partner who never gets tired and never charges you for coffee.

**[PAUSE]**

**MAL:** Okay, confession time. The biggest mistake I made—and most beginners make—is treating AI like a one-shot deal. You ask once, you get an answer, you move on.

Wrong.

AI gets *better* when you push back. Ask a follow-up. Say, "That's interesting, but dig deeper into point two." Refine it. It's an iterative process, not a vending machine. The first draft is rarely your best draft.

**[UPBEAT SOUND]**

**MAL:** Here's your practice exercise for this week: Pick something you actually need—a resume, an email, a summary of something you read. Write a **bad prompt** for it. Just ask plainly. Then write a **good one** with role, specificity, and personality. Compare the outputs.

I guarantee the second one is better. You'll see it instantly.

**[TRANSITION]**

**MAL:** Finally—and this is critical—**always evaluate what the AI gives you**. Don't just copy-paste it into the world like it's gospel. Read it aloud. Does it sound like *you*? Does it actually say what you need? Does it have any weird factual claims that sound plausible but might be totally made up?

If something feels off, it probably is. Your gut is better at spotting AI weirdness than you think.

**[MUSIC BUILDS SLIGHTLY]**

**MAL:** So, if you're here, you're clearly interested in actually *using* AI instead of just watching TikToks about it. That's cool. Subscribe to *I am GPTed* so you don't miss next week when we talk about why your AI keeps writing like a robot—and how to fix it.

Thanks for listening. This has been a **Quiet Please** production. Head over to **quietplease.ai** to learn more.

Now go prompt something. Badly, then well. That's the whole game.

**[OUTRO MUSIC FADES]**

For more...