You're not alone if you think prompting feels like another job.

If you tried AI and got bad responses, it’s not because the tools don’t work.

It’s because you were never shown how to prompt well.

You know how to delegate to people. This is just learning how to delegate to software.

But here’s the truth: The prompt usually matters more than the model.

Let’s fix this.

The D.E.F.I.N.E. Prompt System

Use this framework to write high-performing prompts—fast. It works across any large language model: ChatGPT, Claude, Gemini, you name it.

Each letter stands for a critical ingredient:

D – Describe the Role: Who should the model act like?

"You are a senior consultant with expertise in nonprofit leadership transitions."

E – Explain the End Goal. What’s the final product?

“Help me create a 1-page summary for my board.”

F – Focus the Context: Give the most relevant background, not a data dump.

“Here’s what’s changing in our org, and what we’ve shared so far...”

I – Include Inputs: Paste examples, raw notes, or materials. Don’t make the AI guess.

“Use this bulleted list from our last update and this short quote.”

N – Narrow the Style & Format. Be clear on tone, length, and format.

“Write in a calm, strategic voice. Keep it under 300 words. Use bullet points”

E – Examples of Excellence. If you have a great past version, share it. If not, describe what “good” looks like.

“Here’s what I wrote last time. I loved the summary, match that clarity.”

Your 3-Step Workflow

Step 1 – Run D.E.F.I.N.E. on One Real Task

Pick a task from your friction list (from last week’s issue):

• Weekly email
• Meeting prep doc
• LinkedIn update
• Performance feedback draft

Write a D.E.F.I.N.E. prompt. Paste it. Review the draft. Don’t overthink it.

Pro Tip: Use a launcher tool like Raycast, Text Blaze, or Alfred (Mac) to save your best prompts and reuse them with a few keystrokes.

Step 2 – Use the R.I.V. Loop to Iterate

When the first draft is 70% there, run it through this micro-loop:

• Review: What’s missing, off, or vague?
• Iterate: Adjust just one variable (e.g., tone, structure, or data)
• Validate: Rerun and compare. Does it hit the mark?

2–3 rounds are usually all you need to go from “ok” to “done.”

Step 3 – Run a 60-Second Output QA

Before hitting send or paste:

• Factual accuracy: spot-check two claims
• Tone match: Does it sound like you?
• Risk scan: no sensitive data, no cultural red flags
• Clear CTA or next step included

Safety Check: Prompting with Confidence

This is your company’s and your reputation; ensure you feel good about the output.

Here’s how to stay safe:

• Never paste confidential client info into public AI tools
• Use secure environments (enterprise versions or browser plugins with privacy guarantees)
• Delete chats or outputs that include sensitive internal references

If your team shares outputs, ensure that metadata is scrubbed before it is forwarded or exported.

Make It Measurable

Track how long that task took before using AI.

Now, time it after using D.E.F.I.N.E.

Determine your time savings, revisit your inputs, or ask the model for alternatives if you aren't saving significant time.

This is how you start building real ROI.

Try This Week’s Challenge

Pick one real, high-friction task this week.

Run it through D.E.F.I.N.E. Use the RIV loop once. Track your time savings.

Then save the final prompt + output in a shared folder or prompt library.

Though it may feel like busywork, it isn't, it’s system building.

Final Thought

Most people speak like they talk: unclear, unstructured, and too fast.

But you’re not like most people (I just know :).

The more you treat AI like a high-level collaborator, one that needs smart, focused instructions, the better the output.

You don’t need more tech tutorials. You need a system that works for you, respects your time, and reduces your workload.

This is how you stop “playing with AI” and start succeeding with it.

‍