Welcome back to *I am GPTed*, the podcast where I, Mal – your resident Misfit Master of AI and lifelong subscriber to the “Try Everything at Least Three Times Before Admitting You’re Wrong” newsletter – take you through AI topics without the buzzwords, gatekeeping, or the vague promise that artificial intelligence will bring you inner peace or cook you breakfast.

Today we’re diving into one of my favorite prompting techniques: **role prompting**. That’s right—giving your AI a job title so it actually behaves like it knows what it’s talking about. Think of it like asking your friend Kevin for tax advice… unless you tell him to pretend he’s an accountant, you’re just going to end up with “Have you tried crypto?” as the answer.

**Let’s get practical. Here’s my disastrous “before” example:**

> “Write a summary of this article.”

You’ll get a summary, sure—bland, flavorless, probably lifted straight from the middle of the Wikipedia sandwich tray. Now, here’s the “after,” with a little role-based magic and plain instructions:

> “You are a science writer for a popular magazine. Summarize this article in a way that’s engaging for readers with no scientific background. Highlight why this topic matters today.”

Suddenly, you’re reading something with a pulse, and nobody needs a PhD to follow along. According to the Prompt Engineering Guide, this “role prompting” helps steer the AI’s personality and expertise, and when you tie it to your actual goals—engagement, clarity, not terrifying your readers with jargon—it performs way better than default requests.

**Practical use case time:** Let’s say you’re swamped at work, and your boss wants you to draft a customer-facing FAQ. Instead of wrestling with writer’s block or recycling dusty old templates, prompt AI like this:

> “Act as a customer support specialist for our small business. Create friendly, concise FAQs based on our products and recent customer emails.”

Suddenly your FAQ isn’t just functional; it’s in the right tone, sounds human, and actually helps people. Oh, and you can take that caffeine break you were definitely not going to take anyway.

**Now here’s the mistake I made (semi-monthly, in case you’re tracking):** I used to ask AI for “concise meeting notes” and just…copy-pasted its first try into an email. Spoiler: Half the time it missed the big decisions or mispronounced people’s names in text (don’t ask). The fix? Always review, rephrase where needed, and—my secret—ask AI to critique its own work first: “What’s missing from these notes? What would make them clearer?” That simple ask catches most errors before I embarrass myself *again*.

**Want to practice? Try this exercise:** Pick a simple task—summarize your weekend. First, prompt AI: “Summarize my weekend.” Then change it to: “Act as my witty friend. Summarize my weekend in three funny sentences, focusing on anything I did that was regrettable or entertaining.” Notice the difference? Now you’re thinking like a prompt pro.

**Before I go, here’s a rapid-fire tip:** If AI coughs up a response that sounds weird or half-baked, ask for another version with feedback: “Try again, but be more specific and make it shorter.” Iterating and being picky with your requests is not “being mean to the robots”; it’s essential for quality results. You wouldn’t accept your own first draft—or your first pancake—so why settle with AI?

Alright, time for Mal’s Minute of Humility: When I first tried role prompting, I accidentally told my AI to “act as an enthusiastic cat.” Let’s just say the resulting tech article involved a lot of purring, and very little substance. Lesson learned: be specific, and maybe stick to roles that pay taxes.

Don’t forget to subscribe to *I am GPTed* so you never miss another episode of AI know-how, sarcasm, or the latest in self-inflicted learning disasters. Thanks for tuning in—this has been a Quiet Please production. For more, head to quietplease.ai.

Until next time, remember: prompt responsibly, and double-check before sending.

This content was created in partnership and with the help of Artificial Intelligence AI