This split is what makes AI work

I just asked Claude what VP, my book site, has managed to do today without me even logging in.

Just today it's...

... added 165 books, each with their own page
... discovered 68 new deals, 87 Kindle Unlimited titles, and 55 resurged deals
... written 21,360 words so each page is unique
... added a "buy now / wait for a better deal" verdict to every book
... published 2 carousels of 10 images across 4 platforms
... published 2 thirty-second reels across 3 platforms
... created and sent 1 email to the list with the new offers
... refreshed the prices for 639 books with today's up-to-date info

All done without me touching a thing (I've been watching World Cup highlights), for a total cost of about $4. That's a full growth team running the marketing for less than $5 a day.

The big question for you is, what can you have AI do for your business so you can actually work on growing the revenue?

I'd love to tell you this is all possible because I'm some sort of AI savant.

But I'm not. When I look at some of the content others are publishing in the space, they have a much deeper understanding of how AI works and all the little details.

And yet, not one of those people has done anything like this.

The reason I can do it with my basic understanding of LLMs is one simple rule I follow every time I build a system.

I call it 10/80/10.

Nail this and you can have AI doing complex tasks, predictably, for pennies.

The initial 10%. The part that dictates overall success.

This is human led.

You are better than AI at planning, insight, and strategy, so everything has to begin with you.

This first 10% is about understanding what's gonna move the needle, and the most effective ways to implement that with AI.

The meaty 80%. The part that saves you all the time.

This is where AI does the bulk of the work.

Now, you can't just say "here's what I think, go and do it all please", despite what the AI gurus would have you believe. AI ain't that smart. Trust me, I've tried.

To make sure the AI works as intended, you have to limit everything it does. Think of it as an assembly line. It needs...

... consistent input that does not change across each run
... hardline SOPs it cannot deviate from
... proven templates where it can only adapt what's in them, never decide "how to make it better"

You have to remove the creativity from it. Do that and it works without issue.

The final 10%. The part that ensures quality.

This is you checking what it's done is good enough, and spotting how it can be improved.

That feedback goes back into the first 10% and the 80%, so every run gets better.

This rule is why my systems run every single day without drama. It's not about prompts or small tweaks. It's this rule, and the chassis underneath it that guarantees its success.

And here's the thing. The rule is easy to understand. Building the chassis, the locked SOPs, the templates, the consistent inputs, is the part everyone skips, because starting from a blank page is miserable.

It's also exactly why most people's AI drifts. When the AI doesn't know where to look for rules, it makes them up.

So I've done that part for you.

I've packaged the framework I use as the chassis for every system I run, including the one that did that $4 day at the top of this email, into a launch edition you can download and use today.

If you want the foundation my growth team stands on, it's yours. Free, and yours whether you ever buy a thing from me or not.

​→ Grab the AI Foundation Framework (launch edition) here [LINK]​

Tomorrow, I'll show you a full day of my systems running, so you can watch the 80% do its thing.

Speak soon,

Pete "remove the creativity" Boyle