How are you measuring AI agent ROI in a way that finance actually believes?

The hardest conversation in enterprise AI right now isn't with engineering. It's with finance.

The pattern:

- Engineering says 'agent saved 40% of ticket time'

- Finance asks 'so we can reduce headcount by 40%?'

- Engineering walks it back, finance loses trust

- Project gets reclassified as 'experimental'

What's the ROI framing that's actually worked for you when presenting to finance or the CFO?

Hard cost savings, productivity gains, capacity expansion, something else?"