Per-execution cost, measured
Actual provider usage is recorded per execution with the pricing snapshot in force at the time, so cost is attributable rather than apportioned.
If you resell AI features, your cost of goods is a routing decision. Planverity makes that decision explicit and records what it actually cost.
Blended monthly provider invoices cannot tell you which customer, feature or workload consumed the spend — so pricing is guesswork and a single heavy account can quietly erase margin.
Actual provider usage is recorded per execution with the pricing snapshot in force at the time, so cost is attributable rather than apportioned.
A per-key daily cap bounds any single integration. The important half of a spend cap is usually not the organisation total but the blast radius of one credential.
The comparison baseline is captured at planning time from candidates that actually passed policy — claiming savings against an alternative your own policy would have refused is not a saving.
Cost is attributable per execution, per key and per project. Two real limits: project attribution only populates where a key was bound to a project, so an unbound key still rolls up to the organisation alone; and per-user attribution does not exist at all — an execution is made by a credential, not a person.