Ignore 90% of your AI ideas

The fastest way to get results from AI is to ignore 90% of your ideas for it.

That sounds backwards when every week brings a new tool and a new thing you could build. But a business with ten AI projects on the go usually has none of them finished.

I learned this the hard way in my first business. We spent years chasing great ideas: new systems, new processes, a new team structure, new products and new offers. None of it got finished, so we never got the full benefit of anything we started.

Every business has a bottleneck. It's the one constraint holding everything else back, and the rest of the business has to squeeze through it.

It's rarely obvious and rarely exciting, which is why great ideas so seldom fix it. Fix the bottleneck, though, and a lot of your other problems start sorting themselves out.

AI makes this easier to get wrong. You can now start ten shiny projects in a week, and each one feels like progress. Being able to build something is a poor reason to build it.

Elon Musk uses a five-step process at Tesla and SpaceX. Question the requirement, delete what you can, simplify, speed it up, and only then automate. He's said that on the Model 3 production line he ran it backwards and automated first.

So before you start any AI project, I'd ask three questions.

What's the one number capping your growth right now?

Who decided that part of the business has to work the way it does?

What happens if you delete it?

AI goes on whatever survives those three questions. Everything else is the 90% you ignore.