Token spend sliced by user, by model and by business case — and then set against the value each case is meant to deliver. Spend nobody can attribute to a business case is the first thing this page makes visible.
Value per completed business outcome, not per call: the owner-signed baseline (how long the work takes a human with no AI) minus the review and rework time actually measured. Test runs and failed outcomes are excluded from value and counted as waste.
A planning tool, not a measurement: value here comes from minutes-per-call assumptions you can slide. For a figure to put in front of a CFO, use Measured value above — it is per outcome and built on a signed-off baseline.