Master Your AI Conversations: The Role, Goal, Constraints Framework That Actually Works

[Intro music fades in] MAL: You’re listening to **“I Am GPTed”** – the show where we turn buzzwords into actual useful stuff, and I pretend I have my life together by talking about AI. I’m **Mal, the Misfit Master of AI**. Master mostly because I’ve broken these tools in every possible way, and lived to tell you what *not* to do. Today we’re doing five things: - One prompting technique that instantly improves your results - A practical use case you probably haven’t tried - A super common beginner mistake – that I also made, repeatedly - A tiny exercise to build your AI “conversation muscles” - And a simple way to judge and improve what the AI gives you Let’s de-hype this thing and make it useful. --- MAL: First up: **one prompting technique** that changes everything: **“Role + Goal + Constraints + Example.”** Most people type: “Write an email to my boss about a late report.” That’s like walking into a restaurant and yelling “FOOD.” You’ll get *something*, but you might not like it. Here’s the **before**: “Write an email to my boss about a late report.” You’ll probably get a stiff, formal robot memo that sounds like your boss’s boss’s lawyer wrote it. Here’s the **after** with Role + Goal + Constraints + Example: “Act as a friendly but professional office worker. Goal: Write a short email to my boss explaining my project report will be 1 day late. Constraints: 100 words max, no big corporate buzzwords, sound human and accountable. Example of my tone: ‘Hey Sarah, quick heads-up – running a bit behind but I’ve got a plan to catch up.’ Now write the email.” See the difference? You’re not begging a magic box. You’re **giving instructions to a very literal intern**. --- MAL: Next: **a practical use case** you probably aren’t using enough – **“AI as your boring-life script doctor.”** Not strategy. Not billion-dollar business plans. Just… the annoying stuff: - That awkward message to a client you’ve been avoiding - The “no” email when someone asks for a discount - The “hey, can we move this meeting?” without sounding flaky Prompt it like this: “Act as my communication assistant. Goal: Turn this messy draft into a clear, kind message. Constraints: Keep it under 120 words, maintain my casual tone, don’t over-apologize. Here’s my draft: [paste your ugly message]. Improve it, then briefly explain what you changed and why.” Now AI isn’t replacing you. It’s **editing you on fast-forward**. --- MAL: Let’s talk about a **common beginner mistake**: Treating the first answer like it’s holy scripture. I did this. First time I used an AI model, it gave me a wildly confident, beautifully written answer. It was also… impressively wrong. Like, “don’t let this thing do your taxes” wrong. Here’s how to avoid my shame: 1. Assume the **first answer is a draft**, not the final. 2. Ask follow-ups like: - “Explain your reasoning step-by-step.” - “Give me 2 alternative versions with different styles.” - “What might be missing or worth double-checking here?” If it sounds too slick and you did zero thinking, that’s a red flag. AI is a **calculator with opinions**, not an oracle. --- MAL: Now a **simple exercise** to build your AI interaction skills. Do this three days in a row. It takes 10 minutes. Pick **one small task**, like: “Summarize this article and give me 3 action steps,” or “Help me plan a 20-minute study session for tomorrow.” Then follow this 3-step pattern: 1. First prompt: give Role + Goal + Constraints. 2. Second prompt: “Now improve your answer. Be more concise and prioritize what a beginner would need first.” 3. Third prompt: “What questions should *I* ask you next time to get an even better answer?” You’re training **yourself** how to think in prompts, and you’re training the model how to work with you. Reps, not magic. --- MAL: Finally, a **tip for evaluating and improving AI-generated content**: Use the **F.A.C.T. check**: - **F – Fit**: Does it fit your audience and purpose? Ask: “Rewrite this for [my boss / a 10-year-old / a non-technical client].” - **A – Accuracy**: Are facts correct and current? You still need to check numbers, names, dates, and any strong claims elsewhere. If it sounds very sure, you should be very suspicious. - **C – Clarity**: Is it easy to understand? Ask: “Simplify this by 30%, cut jargon, keep meaning.” - **T – Tone**: Does it sound like *you*? Paste a sample of something you’ve written and say: “Match this tone: [paste sample]. Now rewrite the answer in that style.” If it fails any part of F.A.C.T., you don’t throw it away – you **iterate** and fix it. --- MAL: That’s it for today’s episode of **“I Am GPTed”** – where we use AI like a tool, not a religion. If this helped you boss your AI around a little better, **subscribe to the podcast** so you don’t miss future episodes. **Thanks for listening**, and for admitting with me that we sometimes let the robot be smarter than it should be. This has been a **Quiet Please** production. To learn more, head over to **quietplease dot ai**. For more check out https://www.quietperiodplease.com/ and for some great deals go to https://amzn.to/4nidg0P