The 80-20 of AI Prompt Writing

Writing high-performing prompts is set to become an indispensable skill for knowledge workers.

Especially for product leaders, it will be a must-have capability.

Over the past 48 hours, I delved into the vast array of resources on prompting available.

My objective? To identify the common elements of high-performing prompts.

After an additional 24 hours of testing everything I’ve learned, I’ve summarised my findings in what I deem as the “80-20” principle of prompting.

Introducing the R.O.D.E.S prompt framework.

Here’s a breakdown:

1. Role (R)

Start by defining the persona you want the AI to take on. The more details the better.

Example:

You are a world-class twitter ghostwriter who writes viral tweets.

2. Objective (O)

Clearly state the goal of the task. Be as specific as possible.

Example:

I want you to write a viral tweet on a topic that I will provide you.

3. Details (D)

Provide specific requirements and constraints for your request, including the exact steps you want it to follow.

Example:

The viral tweet should meet the following requirements:

• It should be no longer than 280 characters.

• It should use powerful adjectives i.e. relentless, mind-blowing etc.

• It should not have hashtags or emoticons etc.

4. Examples (E)

Offer examples of the input and output you desire, allowing the model to learn exactly what you want.

5. Sense Check (S)

Finally, ask the model if it understands your prompt and if it needs any further information to complete the objective.

Below is an example of the framework in action.

And that's it!

That's the 80-20 of prompting in a framework you should remember.

Feel free to give it a try and let me know how it goes.