Office hours: who owns the AI budget

This session started with a data mystery and ended up somewhere much bigger: who in the organization actually gets to control the AI budget.

Why a reservation row can show no subscription name

Two attendees were debugging a customer’s cost data: unused-reservation rows showed a blank subscription name, and it looked wrong. It wasn’t. Under a Microsoft Customer Agreement, reservations cross-charge to an invoice section, not a subscription, and the allocation hierarchy runs billing account, then billing profile, then invoice section, a different shape than the billing account, department, and enrollment account hierarchy under an Enterprise Agreement. Know which billing model you’re on before you chase a data bug: the “problem” can be the architecture working exactly as designed.

Who’s going to own the AI budget

The bigger conversation: AI and token costs are becoming a board-level line item, and nobody has settled who owns that budget. One attendee put it plainly:

AI is very soon going to be consuming as much as your payroll, potentially even more. The question in the C-suite is: who’s going to control that budget?

He backed it up with his own GitHub Copilot bill, which had grown sharply once usage-based billing kicked in. That’s not a hypothetical. Michael Flanakin added the broader point: this isn’t an Azure-specific problem, it’s showing up everywhere teams have adopted metered AI tools.

Two things to watch for, both still unshipped: Azure Cost Management is working on an AKS-level cost export (namespace-level “cube cost” data), and the same attendee’s own next project is pulling Azure API Management’s token-metering data into FinOps hubs so teams can attribute AI spend per subscription and per user.

If your organization hasn’t had the “who owns the AI budget” conversation yet, it’s coming. Office hours is where you’ll hear how other teams are getting ahead of it, every other Wednesday.