AI will change the world. But maybe in different ways than we expect. I personally use AI for isolated tasks like ideation, rewriting / formatting, and as a better form of search. But I've struggled to get good, consistent results across more generalized use cases.

Rick Mulready is a podcast and YouTube host focused on building AI automations to improve business processes. Over the years he's helped thousands of creators build effective business systems and improve them with targeted AI automations.

In this post we'll share Rick's 4 essential components for crafting AI prompts that get accurate results consistently.

Rick shared these tips on episode 759 of the Smart Passive Income podcast.

4 Essential Components for Effective AI Prompts

Set Perspective: Setting perspective allows the AI to take on the role of the agent you're looking for. This allows it to give better results - pulling on the context, knowledge, and perspective of the given role. For example - you can imagine that "world-class business strategist" and "middle school student" may give you vastly different results for the same task.

• "You are a world-class business strategist"
• "You are a junior developer"
• "You are a content strategist for a top 10 tech podcast"

Set Criteria: Setting criteria is crucial for the AI to give reasonable results that fit your expectations. Humans require the exact same things but often can infer these unspoken rules from past context / experience with you and their role. AIs are a bit worse at this as they don't really have a particular kind of experience so it's helpful to give them guardrails to play within.

• "Results must be readable in 1 minute"
• "Answers must be between 500 - 1000 words"
• "Steps must be completable within 15 minutes"

Instructions: Now that you've set the perspective the AI should work from and criteria for it to play in, you can tell it what you want it to do.

• "Write a YouTube script explaining the best ways to prompt AI"
• "Write a tweet sharing the benefits of x product"
• "Give me ideas for a mascot for the F# programming language"

Give Examples: AI is very good at translating from one verson of something into another version of that thing. Providing examples of the format you're expecting and exceptional responses will allow it to better understand what you're looking for so it can bias towards the results you like.

• "Give me example blog post titles in the form of TOPIC, FOR WHO, SO THAT"
• "Give responses in the form of a bulleted list with the top 3 reasons you'd choose each"
• (ideally give an example response in the form you've asked for)

Bonus - Iterate on the prompt: You may get the perfect results in the first pass but often you'll need to iterate on the prompt. Human language is often not the most precise form of data conveyance (hence why we have programming languages) so, just like with a human, things may get lost in translation. Perhaps the instructions were unclear or some piece of information / context was missing. If you aren't getting what you want, consider iterating on your prompt to improve it.

• Change your input prompt
• Ask it what prompt you should use to get it to do x
• Ask it what information it needs to effectively do x

Next

Personally I find today's AIs to be a cool piece of technology. I don't think they're ready to disrupt (or take over) the world quite yet but I do think they've got promise and will only get better over time.

If you liked this post you might also like:

• 3 Areas I'm exploring to build more side projects faster and cheaper in 2024
• Stop wasting your time on Bad Startup Ideas
• My journey from Software Engineer to Entrepreneur and back again