Most people get mediocre results from AI because they treat it like a search engine. Here's a simple framework I teach every team I work with — and it changes everything.

"I tried ChatGPT and it gave me garbage."

I hear some version of this almost every week. A business owner types something like "write me an email" into ChatGPT, gets back something generic and robotic, and concludes AI isn't ready yet.

But here's the thing — that's like walking into a restaurant, saying "bring me food," and being disappointed when the waiter doesn't know you wanted the salmon. AI needs context. It needs specifics. And once you learn how to provide that, the results go from "meh" to "wait, it can do that?"

The prompting framework I'm about to share is the single highest-impact thing I teach. People go from frustrated to productive in about 15 minutes.

What is the R.O.L.E. Framework for AI prompts?

R.O.L.E. is a four-part structure for writing prompts that get consistently useful results. It works with ChatGPT, Claude, Gemini, or any other AI tool. Here's the breakdown:

R — Role

Tell the AI who it should be. This sets the expertise level and perspective. Think of it as hiring a specialist for 30 seconds.

"You are an experienced property manager who specializes in multi-unit residential buildings in Central California."

O — Objective

State exactly what you need the AI to produce. The more specific your ask, the more useful the output. Don't leave room for guesswork.

"Write a move-in welcome letter for new tenants at our Bakersfield apartment complex that covers parking rules, maintenance request procedures, and community guidelines."

L — Length & Format

Specify how you want the output structured. Word count, bullet points vs. paragraphs, tone, reading level — the more constraints you give, the less editing you do afterwards.

"Keep it under 300 words. Use a friendly but professional tone. Include a bulleted section for key rules. Write at an 8th grade reading level."

E — Examples

Give the AI something to reference. This is the most underused part of prompting and the one that makes the biggest difference. Paste in a previous email, a competitor's copy, or anything that represents what "good" looks like to you.

"Here's a welcome letter I wrote last year that tenants liked. Match this tone: [paste example]"

What does a bad AI prompt look like vs. a good one?

Let me show you a side-by-side comparison. Same task, completely different results:

Bad prompt:

"Write me an email to a customer."

Result: A generic, bland email that sounds like it was written by a robot. You'd spend more time fixing it than writing from scratch.

Good prompt (using R.O.L.E.):

"You are a friendly customer service representative for a Bakersfield HVAC company that's been in business for 15 years. Write a follow-up email to a homeowner whose AC unit we repaired last week. Check in on whether the issue is resolved, mention our seasonal maintenance plan (not pushy, just informational), and keep it under 150 words. Use a warm, conversational tone — like you're checking in on a neighbor, not selling something."

Result: A natural-sounding email you can send with minimal edits. Takes 30 seconds instead of 15 minutes.

The first prompt gives AI nothing to work with. The second tells it who to be, what to write, how long to make it, and what tone to use. That specificity is the entire difference between "AI is useless" and "AI just saved me an hour."

5 AI prompt templates you can use today

Here are five prompt patterns I give to every team I train. Copy them, customize the brackets, and start using them immediately:

1. The Meeting Summarizer

"Summarize the following meeting notes into 5 key action items. For each item, include who's responsible and a suggested deadline. Flag anything that seems urgent: [paste notes]"

When to use it: After any team meeting to create instant accountability.

2. The Email Polisher

"Rewrite this email to sound more professional but still friendly. Keep the same information but improve clarity and flow. Fix any grammar issues: [paste draft]"

When to use it: Before sending any important email to a client, prospect, or partner.

3. The Content Brainstormer

"Give me 10 social media post ideas for a [your industry] business in [your city]. Focus on topics that demonstrate expertise, build trust with local customers, and could drive people to our website. Include a mix of educational, behind- the-scenes, and promotional posts."

When to use it: Weekly content planning sessions when you're staring at a blank page.

4. The Feedback Analyzer

"Analyze this customer feedback and identify the top 3 recurring complaints and top 3 recurring compliments. For each complaint, suggest one specific action we could take to address it. Be direct: [paste feedback]"

When to use it: Monthly review of customer reviews, surveys, or support tickets.

5. The SOP Builder

"Create a step-by-step onboarding checklist for a new [role] at a [industry] company with [team size]. Include milestones for the first day, first week, and first month. Be specific enough that anyone could follow it without extra explanation."

When to use it: Anytime you need to document a process or train someone new.

The one prompting tip most people miss

Here's the thing nobody tells you: the best prompt writers don't write one perfect prompt. They have a conversation.

You send your first prompt, read the response, and then refine. "That's good but make it shorter." "Now rewrite it for someone who's never heard of our company." "Add a section about our same-day service guarantee." "Make the tone less formal."

Think of AI like a smart collaborator who needs direction, not a vending machine where you put in a query and get one fixed output. The back-and-forth is where the real magic happens. I regularly go 4-5 messages deep before I get exactly what I want — and the final result is always better than anything I could have gotten from a single prompt, no matter how well- crafted.

Frequently asked questions about AI prompts

Does the R.O.L.E. framework work with all AI tools?

Yes. R.O.L.E. works with ChatGPT, Claude, Gemini, Copilot, and any other conversational AI tool. The principle is universal: more context and specificity always produces better output, regardless of the platform.

How long should an AI prompt be?

There's no perfect length, but most effective business prompts are 3-6 sentences. That's enough to include role, objective, format constraints, and tone direction. Don't worry about being "too long" — AI handles detailed instructions much better than vague ones.

What if the AI gives a bad response?

Don't start over. Tell the AI what was wrong. "That was too formal," or "Focus more on the cost savings," or "Make it half the length." Iterating is faster and produces better results than rewriting your prompt from scratch.

Can I save and reuse prompts?

Absolutely, and you should. Build a document with your 5-10 most-used prompts. Swap out the specifics each time but keep the structure. This turns prompt writing from a creative exercise into a repeatable system.

How to Write AI Prompts That Actually Work

Key Takeaways