Your AI cost report shouldn't wait a quarter for plumbing

Gartner says roughly 11% of organisations are entirely unaware of what their function spent on AI in 2025. Not a rough guess. No idea at all.

I get why. We're in the wild phase. Adoption is surging, usage swings month to month, and everyone is trying three tools to find the one that sticks. That's healthy. It's also why cost is a 2026 conversation, and probably a 2027 one too.

You don't need a data platform to start. I built a simple workflow that pulled usage from each tool's API and checked it against a minimum. The policy was one base coding agent per person. It sent usage warnings and regular updates, and flagged licences for removal when people were stacking several tools, or not using one at all, like someone on extended leave. The checks were plain deterministic rules. AI only did the boring bit, normalising each vendor's usage figures into one report.

Nothing clever. Just visibility.

And that's the point. FinOps, the habit of knowing what your tech actually costs, isn't an AI thing. It's total cost of ownership, and you'd want that for any tool, AI or not.

In a few years this settles down. The infrastructure beds in, the spikes level out, and AI becomes normal operating overhead. Nobody will care much then.

But if you're not collecting the numbers now, you'll have nothing to baseline against later.

Start ugly. Start this month.