What NASA's moonshot teaches us about using AI

The Economist published an article recently suggesting AI engineers should learn from mechanical engineers.

As an engineer myself, it rang true.

But it didn't go deep enough.

Early in my career, I built composite airplane engine parts.

Later, I worked with some of the original NASA moonshot team.

They told me something that stuck with me.

In the beginning, NASA couldn't manufacture parts at the tolerances required for the moon mission.

So they changed the approach.

Build as precisely as possible.

Then test everything.

Find the 2% that actually worked under those conditions.

That 2% went to the moon.

The rest stayed on Earth.

They didn't need perfection.

They needed reliability under pressure.

Here's why this matters for AI:

Most people write one prompt and hope it works.

That's like building one part and launching it into space.

Better approach: test variations, find what's reliable, use that.

NASA didn't guess their way to the moon.

They tested their way there.

Same principle applies to building AI workflows in your business.

You don't need the perfect prompt on the first try.

You need a process that finds what works reliably.

That's what I teach.

Not AI theory.

Not the latest tools.

How to test, refine, and build workflows that hold up under real business conditions.

Test. Refine. Find what's reliable. Ship it.