1. Which apps generate this spend?
A single total hides the story. Ask for spend by app, so each team owns its number.
2. What does one unit of work cost?
Cost per support conversation, per document or per generated page is easier to forecast than a monthly total.
3. What happens to the bill if usage doubles?
Usage-based costs scale with success. Ask what the bill looks like at twice and ten times today's volume.
4. Who can create new AI keys?
Every key is a way to spend money. Know who can create them, and which app each one belongs to.
5. How do we know a saving is real?
Any claim of lower cost should come with a baseline: what the same work would have cost before. Without a baseline, a lower bill might only mean lower usage.
Put it into practice
Ask each team for the four figures above once a month. Within a quarter you will have a clear picture of what AI costs and what it delivers.