[Theme music fades in, then under]

You’re listening to “I Am GPTed,” the show where we take artificial intelligence, remove the artificial confidence, and see what’s actually useful underneath.

I’m Mal, the Misfit Master of AI. I break this stuff down in plain English, with just enough sarcasm to keep us all awake.

Today we’re talking about one simple prompting technique that makes every AI you use—ChatGPT, Claude, Gemini, Grok, whatever—instantly less useless: **Show, then Tell.**

Most people just *tell* the AI what they want. Pros **show an example first**, then tell it what to do.

Here’s the “before” prompt:

> “Write a professional email to a client about a project delay.”

The AI will spit out something that sounds like it ate a corporate handbook and is now deeply ashamed of itself.

Now the “after” prompt using Show, then Tell:

> “Here’s an example of the tone and style I like:
> ‘Hey Sarah, quick heads up on the timeline. We’ve hit a snag, but here’s what we’re doing about it…’
>
> Using that same friendly, honest tone, write an email to a client about a one-week project delay because a key supplier missed their deadline. Keep it under 150 words.”

Same task, totally different output. You gave:
- an **example**
- the **tone**
- the **reason**
- a **word limit**

You showed, then told. The AI finally has some guardrails and can stop cosplaying as a 1990s fax machine.

Let’s move to a practical use case you probably haven’t tried: **using AI as your “meeting translator.”**

Next time you get a messy meeting transcript or a wall-of-text notes doc, paste it into your AI and say:

> “You are my meeting translator.
> 1) Summarize the discussion in 5 bullet points.
> 2) List action items with owner and due date.
> 3) Rewrite any vague tasks so a new hire would understand them.”

Suddenly that 60‑minute chaos call becomes a clear to‑do list. No MBA required. No buzzwords harmed.

Now, a common beginner mistake—one I absolutely made: **asking once, accepting whatever comes out.**

I used to type a vague prompt, get a “meh” answer, and think, “Guess AI just isn’t that good.”
No, Mal. **You** just weren’t that good.

Treat the first answer as a **rough draft, not a verdict.** Follow up with:

- “Make this shorter and more direct.”
- “Add two concrete examples.”
- “Rewrite this for a 12‑year‑old.”
- “Give me three alternative versions with different tones.”

The magic isn’t in the first prompt. It’s in the *back‑and‑forth*.

Here’s a simple exercise to build that skill:

1. Pick one small task: a text, email, social post, or explanation.
2. Ask the AI to do it in your default lazy way.
3. Then force yourself to ask **three follow‑up prompts**:
- one to change tone
- one to change length
- one to add or remove detail
4. Compare all four versions and pick the best parts.

Do that daily for a week and you’ll be better than 90% of people using these tools. Low bar, yes. Still true.

Finally, how do you **evaluate and improve** AI content so you’re not just copy‑pasting robot noise?

Use this quick checklist:

- **Clear?** Can a non‑expert understand it on the first read? If not: “Simplify this, remove jargon.”
- **Concrete?** Does it use examples, steps, or specifics? If not: “Add 2–3 practical examples.”
- **Correct?** Sense‑check facts against your own knowledge or a quick search. If shaky: “List your assumptions and uncertainties.”
- **Concise?** If it feels like a TED Talk that won’t end: “Cut this by 50% without losing key points.”

You’re not just consuming AI output—you’re **editing** it. Think of the AI...