[Intro music fades in, then under]

Hey, it’s Mal – the Misfit Master of AI – and this is “I Am GPTed,” the show where we turn buzzword soup into something you can actually use… like lunch. A weird, digital lunch.

Today I’m giving you one simple prompting technique, a sneaky real‑life use case, a mistake I personally keep making, a quick practice exercise, and a fast way to clean up the AI’s mess before you hit send.

Let’s get to it.

---

So, one prompting technique that instantly improves your results: **role plus format plus constraints**.

Translation: tell the AI **who** to be, **what shape** you want the answer in, and **the rules** it has to follow.

Here’s the lazy, “before” version:

> “Explain blockchain.”

Every model on earth will now send you a 700‑word Wikipedia tribute.

Here’s the upgraded “after” version:

> “You are a patient high‑school teacher. Explain blockchain to a 15‑year‑old who hates math. Use a real‑world money analogy, keep it under 150 words, and end with one sentence: ‘If you remember one thing, remember this: …’”

Same topic, totally different vibe. You’ve told it:
- Role: patient high‑school teacher
- Format: short explanation plus one final sentence
- Constraints: teen, hates math, real‑world analogy, 150 words

You can do this in ChatGPT, Claude, Gemini, Grok – they all respond better when you stop mumbling and actually give them a job description.

---

Now, a practical use case beginners usually don’t think about: **being your “second brain” for boring recurring messages.**

Not presentations. Not novels. I’m talking about those awkward, repetitive things:
- “Sorry, I’m declining this meeting but still trying to sound like a team player.”
- “Following up without sounding desperate.”
- “Reminding the client they owe us money… politely.”

Try this:

> “You are my polite but assertive email assistant. Rewrite this follow‑up so it’s friendly, confident, and under 80 words. Keep my tone casual, no corporate clichés. Here’s my draft: [paste your mess].”

You’re not asking the AI to be you. You’re asking it to be your **editor with social skills**.

---

Common beginner mistake time – and yes, I do this too: **asking once and accepting the first answer like it’s sacred scripture.**

I still catch myself doing this:
I type a vague prompt, get a meh answer, sigh, and think, “Guess the AI just isn’t good at this.”

No. I wasn’t good at asking.

Instead of giving up, respond to the AI like this:

> “This is too generic. Make it more specific to [my industry / my situation], add 3 concrete examples, and cut the fluff.”

Or:

> “You missed the part about [X]. Rewrite it and focus mainly on that.”

Treat it like an **iterative conversation**, not a vending machine. If the first answer is bad, that’s not the ending – that’s the first draft.

---

Here’s a simple exercise to build your AI skills – takes five minutes:

1. Pick a tiny task: summarize a page of text, write a short email, or plan a 3‑item shopping list dinner.
2. Write your **first** prompt quickly. Run it.
3. Now write **version two** of the prompt using role + format + constraints. Run that.
4. Compare the two answers and ask:
- What did the better one have that the first prompt didn’t?
- Did I say who it should be? What format I wanted? Any limits?

Do this once a day for a week. You’ll accidentally become “that AI person” in your office, just from being slightly less vague than everyone else.

---

Finally, a tip for evaluating and improving AI‑generated content so you don’t copy‑paste yourself into disaster.
Back to Episodes