Ai prompt Engineering Learnings
AI Prompt Engineering
Last week, MIT released a study stating that 95% of enterprise AI pilots are failing. It reminded me a bit of the dot-com bubble (though I was only 17 at the time). The key difference I see between AI and the dot-com era is that I view AI as a sidekick—not a replacement, and not an entirely new platform (with the possible exception of search—look out, Google).
This summer, I was fortunate to take a college course that helped me refine how I use AI as a sidekick in two key ways:
-
Creating custom ChatGPTs to assist with daily life and improve efficiency.
-
Developing stronger prompt engineering skills.
The second point is the focus of this blog post, and I hope you find it helpful.
A prompt has evolved from a simple run-on sentence to a structured process that positions you as the behavioral designer of the AI assistant you're working with. Here’s the prompt framework I’ve been using:
Role – Define the perspective you're writing from to help frame the problem.
Example: I am a small business owner struggling with X.
Task – Tell the AI what you want it to do based on your role.
Example: Please analyze dataset Y and help me address problem X.
Context – Provide background information or reasoning that explains the situation.
Example: I’m struggling with problem X due to a lack of resources A, B, and C.
Constraint – Set realistic boundaries around time, money, or other limitations.
Example: I can’t invest more than F dollars or H hours to solve this issue.
Goal – Clearly define what success looks like in this scenario.
Example: I’d like three proposed solutions to problem X that stay within the constraints. Please format them as digestible bullet points.
Hope that helps. Thanks for the read.
- Ben
