And what changes first? Is it policy? Is it tooling? Is it training or approvals?

Yeah, no, that's another great question, because these are real conversations that I'm a part of almost every day now in terms of this, what we call shadow AI, right? And the short answer is that once AI starts touching real data, is a part of real decisions, is a part of external facing work, it's no longer experimentation, it's implementation.

And so the biggest signal you're looking at for them, it stopped being experimentation as soon as it touched your real workflows, your real business work.

Now, here's the tricky part.

And this is why it's called in the shadows.

People are already doing that.

And leadership does not know about it.

So what's your minimum evidence standard before something can ship? Is it citations? Is it data lineage, human sign-up, versioning, or audit trails?