3 Approaches for Prompting AI Efficiently and Avoiding Chaos
How I Use Smarter Prompts to Get AI to Do More and Save My Sanity
Yesterday, we partnered with #ChatGPT to pilot our first enterprise AI tool for employees. A big step forward — and a bit of a milestone.
It got me reflecting on how I use AI tools every day to streamline my work. Whether it’s automating tasks or building entire machine learning models, good prompts make all the difference between “spot-on” and “what on earth is this?”
Here are three practical, battle-tested ways to craft better prompts for AI tools like ChatGPT and Claude.
1. Graduated Prompting
When I’m asking the AI to help me understand a new and complex topic, I ease in. Think of it like walking into the ocean — don’t jump straight into the deep end.
I start with prompting the AI to teach me the basic level information. Then I “graduate up” to prompt the AI for more advanced explanations.
I’m a practical learner (I need real-world examples or to try something myself to cement new learning) so I also prompt the AI to give me basic examples to start and then slowly increase the example complexity to more advanced use-cases.
I found this graduated process helps me quickly grasp the basics and steadily move along the learning curve to comprehend topics that would otherwise take weeks or months to learn on my own.
Let’s say I’m tackling global trade dynamics. I might start with:
“Explain the top factors influencing global trade. Keep it simple, and use clear, relatable examples a 10-year-old could follow.”
Once I’ve grasped that, I’d level up:
“Now explain it like I’m 18 and have taken high school economics. Assume I got a solid B — enough to know the basics but with some gaps.”
Each step builds on the last, adding complexity only as I’m ready for it. It’s an iterative process that works wonders for unpacking big topics.
2. Modular Building
Here’s a fun fact about AI: it’s great at solving problems, but ask it for too much at once, and things can go sideways fast.
I’ve been using AI for coding assistance for roughly two years now and have found it tends to hallucinate or forget earlier prompts after 6–10 iterations, depending on the complexity of the ask.
To avoid hallucination while getting coding support on a larger project, I’ve learned to break the prompting asks into smaller modules.
For instance, I might start with a basic ask:
“Write JavaScript that pulls today’s weather forecast from API XYZ in AppScript. Add comments to explain each step.”
Once that’s sorted, I’ll layer on the next request:
“Great! Now adjust the script to save the data into Table XYZ with columns A, B, and C.”
And so on, and so forth.
This modular approach avoids the chaos of a single, all-encompassing prompt that tries to handle everything — data pulls, formatting, scheduling, QA checks, email alerts — you name it. I’ve doun AI performs best when given clear, focused steps.
3. Decision Prompting
One of the underrated powers of AI tools is their ability to help you make better decisions — especially when you’re stuck.
I start by prompting the AI with a lot of context. Spend most of your time here and it’ll pay dividends later.
I’ll prompt the AI with the following inputs for the decision at-hand:
- What’s the overarching context? (put a lot of time here)
- What’s the decision? (be specific)
- What are the options, pros, and cons? (over-index here, the more the better)
- What data is potentially most helpful to decision? (be generous here)
- What biases are you aware of? (take a moment to reflect)
- Why are you struggling? (be honest)
- What decision, specifically, do you hope to make here? (the more specific the better)
Here’s a simple, recent example: I was juggling several side projects, all interesting, all competing for my time.
ChatGPT had helped with each projects, so it had some prior knowledge of the details.
I explained my goals, why I was stuck, and that I‘m hoping to move forward. I prompted the AI with answers to all of the above questions, spending plenty of time giving it as much information as I could think of to help with the decision making process.
To add to these answers, the AI also “knew” that I’m a fan of the Snowball Method for debt management from prior discussions (the Snowball Method suggests paying off your smallest debt first to build motivation and momentum that continues into paying off more and more debt).
I mention this past context of my bias to the Snowball Method because it highlights how AI tools like ChatGPT “get to know” you over time and can use this learning to make recommendations that personalized to you.
Its advice after deliberating over all the prompts and my bias for momentum? Finish the project closest to completion — quick wins free up time for the bigger ones.
Of course, this is a very simple example but it was just the nudge I needed to move forward.
This prompting approach can be applied to simple and complex decisions. It also works across both personal and work-related decisions.
I’ve found AI to offer pragmatic and sensible approaches most of the time, highlighting its future potential for more automated decisioning for businesses in future (more on this coming in a future post!).
Final Thoughts
AI tools are incredibly powerful, but they need structured direction. With clear, thoughtful prompting approaches like what I’ve shared here, AI can massively amplify your productivity and unlock new ways to tackle challenges.
These three strategies — graduated prompting, modular building, and decision prompting — are some of my go-tos for reliable results.
