I want to start this post by giving you a little background.
I recently earned the “AI Architect” certification from Anthropic, and I have a BS in a computer-related field (MIS). I’m not saying this to brag, but because there is a lot of garbage advice floating around about AI right now, and I want you to have at least one person in your feed who actually understands what’s going on under the hood.
If you’re interested in AI— I’m going to show you what actually improves results...and it's not what most people think.
Why AI Makes Mistakes, and How To Fix It
AI usually screws up for two reasons:
- Your prompt is incomplete, or
- Your context is missing or weak
People love to talk about AI making mistakes, but most of the time the model is just guessing because you didn’t give it enough to work with.
If you don’t like the results you’re getting, 90% of the time, this is the fix.
The Most Common Mistake
Most people treat AI like Google.
They type something like:
“Write me a real estate listing for a 3-bedroom home in Phoenix.”
And then they get a generic, boring answer and say:
“AI is not that great and you can tell that it's AI.”
But the AI didn't make a mistake. The AI was trying to answer your question, based on the limited information you gave it. The AI is ignorant! It's very smart, but ignorant. Like a child savant who can play piano by ear, but doesn't understand music theory or how to read.
Bottom Line: You didn’t give the AI anything to work with.
Here is The Fix!
Here’s a trick I use constantly. Before I even run my real prompt, I paste this right above it, in the same window.
Hello GPT! I am going to paste a prompt here, which I intend to give to an AI agent. Do you understand what I am trying to do here? Please re-write the prompt so that it includes all of the stuff I am forgetting. I don’t know anything about anything. Please engineer this prompt for a one-shot success. I know I am missing a lot.
Then I paste my real prompt underneath it. Whatever crappy little prompt that I was going to write.
What happens next is the real magic. The AI doesn’t just “rewrite your prompt"- it upgrades the entire task that you gave it.
It will rewrite your prompt and give it back to you. It will include instructions like, "ask clarifying questions if you are not sure; don't assume anything." The new prompt will tell the AI to structure the output in a way that makes sense; it will include legal and compliance issues, and it will define the tone and demographic.
In other words…
It stops being a lazy assistant and starts acting like a professional. You just had to tell it to be a professional.
Pretty simple, right?
Real Estate Example (this is where it clicks)
Let’s pretend we are asking ChatGPT to help you write a good listing description. Here is a bad prompt:
“Write a listing description for a 3-bedroom home in Boca Raton and here is the address. It's a condo.”
It will do it, but it will write generic, boring content that "looks like AI."
Now watch what happens when you improve context.
You don’t just say “write better.”
You say:
“Here is everything you need to know about the property. Here are my past top-performing listings. Here are a sample of my blog posts and previous listing descriptions, so you understand my tone. Here is my target buyer. Here is what I do NOT want included.... and Here are compliance rules. Please write the listing description and try to sound like me.”
Same AI, but with a completely different result. And that's because you did not allow it to guess.
Create New Projects, Not New Chats
This is where it gets even better. Instead of starting a new chat every time, you can create a “Project"-- which you can do in every public AI.
Think of it like hiring a new assistant and actually training them. Let's say you start a project called Real Estate Marketing.
Inside that project folder, before you ever ask the first question, you are going to upload documents to that folder-- essentially training your employee. You could upload your best listing descriptions, your email templates, your tone and brand voice, your brokerage guidelines, state laws, past social media posts, etc, etc.
Do you really trust AI to search the internet to fill in all of the details that you failed to give it? (The answer is "hell no.") But by giving the AI context, you will get much better results, and almost completely eliminate errors and AI-sounding articles.
Because the AI isn't guessing anymore. You gave it a set of rules and you don't give it room for guessing.
It’s working inside your business model now.
When you setup projects (no more chats!), think about it like hiring a new employee. When you hear that AI is going to take our jobs- they don't mean AI Chat-- they mean projects.
This post got long, but in my next post, I'll tell you how to make AI sound like a human.
Comments(10)