Last week I gave the hubby a prompt.
He needed to clean up an address list before a marketing campaign — the kind of work literally made for robots.
Here’s what I sent him:
Act as a research marketing assistant and review every address on the attached for accuracy. Return as a spreadsheet with a score of 1-10 on your level of confidence. The goal is to ensure that the marketing campaign has the correct targeting addresses.
Three sentences. He got a clean, scored spreadsheet back in minutes.
Now — that prompt isn’t magic. But it’s built on a framework that is.
RAAC.
Role. Ask. Attachments. Context.
Four letters. Every good prompt has all four. Every bad prompt is missing one.
✨ Steal This — The RAAC framework
Next time you sit down to prompt something, don’t just type what you want. Build it like this:
Role — Who is the AI being right now?
His prompt: “Act as a research marketing assistant…”
Ask — What exactly do you want it to do?
His prompt: “Review every address on the attached for accuracy. Return as a spreadsheet with a score of 1-10 on your level of confidence.”
Attachments — What does it need to look at?
His prompt: the address list itself.
Context — What does it need to know that isn’t obvious?
His prompt: “The goal is to ensure that the marketing campaign has the correct targeting addresses.”
That’s it. Four pieces. In any order, but all four.
The reason most prompts fail isn’t that AI is bad. It’s that we type the Ask and skip the other three. So the AI has to guess what role to play, what to look at, and why it matters. And when AI guesses, you get generic.
Try this with one prompt today. Just one. Where you slow down and write all four parts.
You’ll see the difference immediately.
💡 What’s Actually Working
Here’s what I keep coming back to with RAAC.
It’s not really a prompt framework.
It’s a thinking framework.
Watch what happens when you write a real RAAC prompt. The Role makes you pick a perspective. The Ask makes you name a deliverable. The Attachments make you go find the doc you’ve been meaning to organize. The Context makes you say out loud who this is for and why.
By the time you finish writing the prompt, you’ve done 80% of the thinking.
The AI just finishes the sentence.
That’s the part nobody tells you. The people getting real work out of AI aren’t typing better prompts. They’re thinking more clearly before they type.
His address-cleaning prompt works because he knew what he wanted before he opened the chat window. RAAC just made him write it down.
Quick favor
If you try RAAC this week — even badly — hit reply and tell me what you used it for. I read every one, and I’m building the next few issues around what’s actually showing up in your inboxes.
—Jamie
P.S. I’m opening the founding cohort of AI Build Lab in a few weeks. Four weeks, six to eight builders, one task you do every week turned into a system that runs without you. Founding rate, only this once. If ”build it once, stop redoing it” is on your wishlist this year — watch this space.

