𝐂𝐥𝐚𝐫𝐢𝐭𝐲 𝐨𝐯𝐞r 𝐕𝐞𝐥𝐨𝐜𝐢𝐭𝐲

If the game plan to do this thing, whatever it is, whatever it is, is not gonna have the desired impact and is not strategic, you call a timeout and say, listen, we're not doing this. That's right. But we got a deadline.

That's right. Go ahead. Well, here's the interesting thing. AI, now that it's part of the mainstream part of product teams, AI forces this, right? Because AI models start bad. There is no good AI model out of the box. They start wrong before they get right. You train them.

Artificial intelligence models by virtue of them start wrong? Yes. Because you're—

basically training it on data and then you're asking it to make a prediction and you never train it with enough data 'cause you don't know what you're doing.

So you're revising as you get input. We call it an experiment.

It's been called an experiment for decades, right? This is not a new thing, but now we're putting it in companies as their main product and they're like, "No, no, no, March, we gotta get this puppy out the door." And everyone's like, who came from the AI side is like, "This could fail tomorrow. The model could implode."