We have learned a few neat features of basic AI, now let's start learning how to interact with it!
Two people, SAME free ChatGPT. One gets bland junk and says, "AI is overrated."
The other saves an hour before lunch. They are both using the Same tool, no secret upgrade.
The ONLY difference is how they are asking for it.
Today you become the second person. 👇
The reframe I hold in my head for this is AI is not a search box and not a mind reader. It's an incredibly capable helper who does EXACTLY what YOU ask for, and AI cannot read your mind.
So, the whole game is giving AI the same brief you would sending someone to the store.
🙍🏪The stage is set we are sending a neighbor to the store. 🛒
Just by asking our AI to "Get orange juice" makes the AI a talented guesser at what actually comes back. So, what you get is "3 gallons of Tropicana Orange Juice, Lots of Pulp, and a pitcher". The return is a best guess as to what you meant by "get orange juice, and you also got a pitcher to put it in which you did not ask for.
Now, let's ask the same neighbor, but let's provide a better brief. Every AI request we send is a brief, and they work best when they are detailed instructions of exactly what we need. It's a specific errand, that will return great result.
Example: "I need you to go to the store for me, I need you to pick up one gallon of Pulp-Free Tropicana Orange Juice, store brand is ok if they're out, but NO PULP is mandatory, if they have no pulp-free option, don't buy it, just come back and tell me, do not pick up anything else orange juice only." = exactly what you wanted.
THE RECIPE (4 parts of a good brief)
1️⃣ Identity: Who it should be and what lens is it working with you from. "You are my friendly neighbor"
2️⃣ TASK: Exactly what to do, start with a verb. "I need you to go to the store for me"
3️⃣ CONTEXT: The details it needs, and the constraints. " I need you to go to the store for me, I need you to pick up one gallon of Pulp-Free Tropicana Orange Juice, store brand is ok if they're out, but NO PULP is mandatory, if they have no pulp-free option, don't buy it, just come back and tell me, do not pick up anything else."
4️⃣ EXAMPLE: Show AI a sample of the output you want. Notice there's no example for the orange juice; a simple fetch doesn't need one. 😅 Examples earn their keep when you care about the shape of the answer: "write this email like THIS one" or "format it like this list." When you're picturing a specific style or structure, show it. When you just need the thing fetched, skip it. 🍊
A strong brief will bring you what you are looking for, instead of what AI is guessing at what you are asking for. Which will move the needle in your interactions with AI! Enjoy your orange juice! 🍊🧡
4 power moves that multiply results
- 💡 Show it an example of what you are looking for.
- 💡 Tell AI the things you want it to AVOID
- 💡 The pro move: end with "Before you head to the store, ask me any questions you need to do this well." Now AI interviews YOU and fills its own gaps before it even starts.
- 💡 First answer is a draft, not a verdict. Keep going, it's a conversation, not a vending machine.
Here is something you can try today: pick ONE real task, ask it lazily, then ask it with all 4 parts
Then I want you to compare what you get. That gap is the whole skill.
Save the strong version and iterate until its exactly what you want. 🤓
Push back on YOURSELF first, next time you get junk, re-read your prompt before blaming the tool. Most "bad AI answers" are vague asks in disguise. 💪
Post 5 of Uncomplicate AI: A Series for the Rest of Us.
We learn together, we grow together, we win together! 🤓💪🏆
Bas