How to prove an AI contract saved money
A before-and-after method CFOs, controllers and PE operating partners can accept: separate what a renegotiated AI rate saved from usage growth and the vendor's own price changes.
By Keyveve, Inc. · Updated 2026-10-30
You renegotiated an AI contract: a bigger discount, a better tier, fewer seats. Three months later the bill is higher than before. Did the deal fail? Usually not: usage grew. The job is to show both at once.
Why the bill alone can't tell you
AI bills move for three reasons at the same time:
- Usage: more calls, longer prompts, a new feature, a new team on the tool.
- Your rate: what you negotiated.
- List prices: vendors cut, and occasionally raise, their own prices.
Comparing this month's bill with last year's mixes all three. A savings claim built that way won't survive a review.
The method: price and volume, separately
Freeze a baseline before the new terms start: each model's usage over a recent period, valued at the list prices of the time, and what you were billed per dollar of it (your rate). For seat tools, the seats and the price per seat.
Each month after, value the month's usage at the same frozen list prices and split the change:
- Usage change = old rate × (this month's volume − baseline volume)
- What the new rate saved = (old rate − new rate) × this month's volume
- What list prices saved = the part of the new rate that's the vendor's own price change
These three add up to the change in the bill exactly, which is what makes the result checkable. Finance teams know this as a price-volume variance.
Hard savings and avoided cost
When usage grew, the bill went up but would have gone up more without the deal. Show both: the change in the bill against the baseline (what hit profit), and what the new rate saved on this month's usage (including cost avoided on growth). Annualise the monthly rate saving for the run-rate figure an operating partner or board will ask for.
Check the rate is on the bill
A negotiated rate that isn't applied is the most common gap. Each month, compare the rate billed with the rate agreed; anything above it is a billing error to claim back.
Proven savings freezes baselines and produces this report each month, with the check, as a page, a PDF, a CSV or through the API.
Veqen watches what your work and coding agents, your product and your teams spend on AI, names who spent each dollar, and stops the spend nobody on your team made. Run a free audit or read the quickstart.