AI Isn't Magic. It's Just Math.

Every few weeks there's a new demo that's supposed to change everything: an agent that can browse the web, an assistant that can plan your week, a model that can reason its way through a puzzle. They're impressive. They're also mostly not where the money is yet.

If you actually look at the data, the boring use cases are the ones printing money right now. Klarna's customer support AI handled 2.3 million conversations in a single month, roughly the output of 700 full-time agents, and the company projected a $40 million profit boost from it. That's not a chatbot doing something magical; it's a chatbot doing something tedious, at a scale and consistency no human team could sustain. Invoice processing tells the same story from a different angle: the industry average to process an invoice manually sits around $15, and automation drops that to under $2, an immediate 87% reduction in operational cost on a task nobody was ever excited to do by hand.

The pattern across both examples is the same. The businesses winning right now aren't trying to replace their people with something smarter than them. They're using AI to delete the drudgery: the repetitive, rules-based, high-volume work that was never a good use of a skilled person's time in the first place. That's a much less romantic story than "AI agents," but it's the one with actual P&L impact today. The sci-fi version of AI will probably arrive eventually. The math version is already here, and it's already profitable, which is exactly why it's easy to underrate.