I Made It Up.

It was only the second time Andrew had ever spoken on a stage.

He opened by asking how many people in the room had ever been "scared shitless and shaking" on a stage. Then he told them that morning his wife and his fourteen-year-old daughter had both sent him messages to calm his nerves.

So this was not a man with a lot of confidence to spare.

And thirteen minutes in, in front of a room full of founders, he told them about a term that didn’t exist yet.

Because he’d invented it.

Then he asked the room a question.

Someone in the audience answered with one word.

Decision.

"Right?" Andrew said. "The decision."

Six months later, the rest of the industry caught on.

But to understand why Andrew saw it first, you have to go back to the first question everyone asked AI.

In the beginning, everybody was trying to write the perfect prompt.

Andrew was too.

He built more than a hundred custom GPTs. People still use them. When ChatGPT sent out its year-end recap, it put him in the top one percent of all its users.

But the work that taught him the most was the work that broke.

Back when AI voice agents were brand new, Andrew was building them. And he ran into a wall that no prompt could get him over.

Think about how many ways a person can answer a phone.

How many ways they can say let me talk to a human.

Andrew had to account for them. In his words, "you had to map everything. Every decision."

He called it "one of the biggest things that I always struggled with."

He didn’t have a name for it yet.

But he’d just met the problem that would follow him for years.

The hard part was never the words the machine said.

It was everything that had to be decided before it said them.

Then the conversation moved on.

People figured out that a better prompt only got you so far. The machine also needed to know who you were, what you sold and who you sold it to.

That became known as context engineering.

Andrew was already teaching it.

Months before Cabo, on stage at a live event in Miami, he told a room of business owners to stop obsessing over prompts.

What mattered, he told them, was giving the AI "a full picture of you and your business."

And he’d started living that way himself.

He recorded his calls. He pulled his messages off his phone. He fed his conversations, his meetings and his notes to his AI so it could actually know him.

Which is how he noticed the thing nobody was talking about yet.

Two days before Cabo, Andrew was planning his talk out loud into a voice recorder.

He was working through what to teach. Ways to use AI in your business, your family, your health.

Then he added a note to himself.

He kept going.

He brought up Blockbuster and Netflix. The same technology. Very different decisions.

And then he asked the question that turned a prediction into a method.

What if you could look back at a bad decision and work out how you got there?

By the end of the recording, he knew.

Two days later, he walked on stage and gave it a name.

In Cabo, Andrew explained it through someone most of the room knew.

He’d spent about a year and a half as COO to Lead Gen Jay.

And the most important thing he learned from Jay had nothing to do with tools.

Andrew admitted he struggles with that. He wants to do everything. Most founders do.

Jay didn’t.

That was the whole idea.

Every business already runs on thousands of decisions. Who gets a reply. Which lead is worth it. What to build now and what to leave for later.

The people who are best at those calls can’t always explain how they make them.

Andrew believed AI could learn to see it anyway.

Which, if you’ve ever tried to hand off part of your business and watched someone make a call you never would have made, you probably understand.

The information was the same.

The decision wasn’t.

After Cabo, Andrew didn’t let it go.

He brought it up in his own working sessions. When he was mapping out the system that held his life’s information, he wrote that "decision engineering" would be the next frontier "after prompt engineering, context engineering," and that capturing "the difference between the right and wrong decisions is literally what makes the difference between Netflix and Blockbuster."

He brought it up on calls.

And when one of them pushed back with chess, pointing out that a computer always knows the goal is to win, Andrew had an answer most people wouldn’t have thought of.

Sacrifice the piece. Lose the battle on purpose. Make the call that only makes sense if you know what’s really at stake.

That’s what he meant by a decision.

But he was also honest about something.

He could feel the idea. He couldn’t fully hand it to anyone else yet.

Then the world moved.

A new kind of AI model came out. It wasn’t built to chat or write. It was built to do one thing.

Make decisions.

Andrew heard about it and couldn’t contain himself.

He described what it did. You give it a question and the rules around it, and instead of a paragraph, it gives you back a decision and how confident it is.

It was six months after he’d said the words out loud in Cabo.

Now there was a tool built for the exact thing he’d been describing.

Days later, Andrew stood in front of his own community and finally laid it all out.

He started with the honest part.

Then he defined it.

The three questions, in his words, went like this.

Prompt engineering asked: how do I ask?

Context engineering asked: what does it also need to know?

Decision engineering asks what should actually be decided, "under what rules," and "what happens next."

And he’d already put it to work. He’d been trying to work out how to enrich cold email leads without it costing a fortune. He took the thresholds and rules he’d already worked out and handed them to the new model.

Then he told the room why he’d picked the word engineering.

Here is the part Andrew wanted people to understand most.

The new model doesn’t come up with the questions. It doesn’t know which calls matter in your business or what’s at stake in them.

Somebody has to design that.

That’s the shift.

For years, the value was in knowing the right words to type.

Then it was in knowing what to feed the machine.

Now it’s in the judgment. Knowing which decisions your business really runs on, and how the best person you know would make them.

Andrew said it took his brother "screaming in my ear" for him to see how far ahead he was.

He’d been ahead the whole time.

He’d hit the wall building voice agents, when every possible answer had to be mapped by hand.

He’d told a room in Miami to stop worrying about prompts.

He’d stood in Cabo, nervous, on only his second time on a stage, and said the name of something nobody had heard of, because he’d made it up.

And he’d done it by asking the same questions he wanted everyone to ask about their own worst calls.

What did I miss? What didn’t I ask?

Turns out, he hadn’t missed much.

Before the output, there was always a decision.

Andrew just saw it first.