Grant Harvey

You've got to have a clarity in your data landscape.

You've got to understand what truth is so your models can actually, or whatever the hell, can do stuff.

Another great problem is you'll be like, okay, here's your process.

You said you've got this problem, right?

Let's understand your process, right?

What's step one?

Oh, step one, I just talked to Day's team and they do something.

Dave's team is the ultimate black box.

If you don't know what Dave's team does, how the hell can you get someone else to do that?

Or you can train a system to do that.

You've got to know what the hell Dave's team does.

That's right.

You've got to understand what your outputs are, what your inputs are.

You've got to understand your process.

That doesn't necessarily mean mathing out your current process, because another big problem is people are like, oh yeah, my process runs like this, AI-ify it.

Imagine if you'd said to someone in 1900, like, right, I want the best and fastest form of land transport we can get.

They'd be like, here's a really fast horse.

We'll make it faster.

Like the motor car is a different paradigm.

In a similar way, your process might be some Frankenstein's monster of bolt-ons and other things.