For twenty-five years, the way businesses cut risk was numbers. Enough inputs, run a regression, get a decent read on the output. Social media runs on it. Credit scoring runs on it. Feed the machine numbers, it finds the pattern, it tells you what's likely. Deterministic and powerful, but it only ever spoke one language: numbers in, number out.
AI changed the input. What it actually does is turn pieces of words into numbers, do math on them, and find the patterns in how words become sentences, sentences become paragraphs, and paragraphs become arguments. Those patterns were always there. Our brains just can't hold them, because our brains are lossy and built for people and relationships, not for counting.
So now we have two kinds of machine. The old one takes numbers and gives you an output. The new one takes human context, the ideas and concepts and messy stuff, and gives you an output. We have been able to calculate numbers at scale for decades. We have never been able to process concepts at scale. That's the step change, and almost everyone is staring at one machine or the other.
I think the money is in the seam between them. The numeric models know what is statistically likely. The concept models know what a situation means. On their own each is half blind. Put them together and one gives you the odds while the other gives you the story, and the story is usually what you were actually deciding on.
It's also why AI writes code so well, and that's a tell. Code is just a function: clear input, clear output, and the machine reasons its way through the middle. Anywhere you can name the input and the output cleanly, this works. The real question is how much of your business is actually shaped like that, and how much of it just hasn't been written down yet.