The new question for CTOs isn't whether developers should use AI. It's who owns the meter.

For much of the past decade, the economics of enterprise technology have been relatively easy to understand. Software was licensed, cloud infrastructure was monitored and budgets, while rarely small, were broadly predictable. Costs rose and fell, but they did so within familiar boundaries.

Generative AI is changing that.

Unlike a software licence, AI is consumed rather than owned. Every prompt, API call and automated workflow carries a cost. Individually those costs are almost imperceptible. Across hundreds of employees and thousands of daily interactions, they begin to tell a different story.

The technology industry has spent the past two years asking whether businesses should embrace AI. Most have already answered that question. The more pressing challenge is how to manage it.

Who decides which models are used? Who monitors consumption across departments? At what point does experimentation become operational expenditure? And when the monthly bill arrives, whose budget absorbs it?

These aren't theoretical questions. They're beginning to appear in boardrooms alongside more familiar discussions about cyber security, cloud spend and technical debt.

It also marks a subtle change in the role of the CTO. The conversation is no longer centred on choosing the right technology. Increasingly, it's about understanding the economics of that technology once it has been placed in the hands of thousands of people.

That explains why FinOps, once the preserve of cloud infrastructure teams, is evolving into something much broader. Managing AI consumption is becoming as important as managing AI capability. Choosing the most powerful model is one consideration. Choosing the most appropriate one may prove to be the more valuable skill.

There is a temptation to see AI as another software purchase. It isn't. It's a utility. Like electricity, the cost is determined less by whether you have access to it than by how, where and how often you use it.

The organisations that thrive over the next few years may not be those deploying the largest number of AI tools. They'll be the ones that understand what each interaction is worth and who, ultimately, is responsible for the meter.