Cost per accepted outcome
The total cost of getting one AI-assisted result that is actually accepted and used, counting model spend, retries, human review, exceptions and recovery — a truer measure than cost per call.
Cost per accepted outcome divides everything spent on an AI-assisted workflow by the number of results that were accepted and used. It answers "what does one good result actually cost?", where cost per call or per token only answers "what did the model charge?".
The total in the numerator includes model and inference spend, retries and tool calls, human review time, exceptions and escalations, failure and recovery work, and governance effort. The denominator is dependable outcomes, not attempts.
Why it matters
When the model gets cheaper, per-call cost falls, but per-outcome cost may not. If a cheaper model needs more retries or more review, the saving moves into other budgets. Outcome-based cost metrics are already emerging; what they need is an extra discipline: the cost-migration test, which asks where the cost went when it moved.
Used with a cost-migration ledger, the metric makes total cost of ownership visible and attributable. It also lets you compare a change in model, route or scope on the thing that matters: the cost of a result someone can depend on.
Read more in A cheaper AI model can move the cost instead of removing it.
Related terms
Total cost of ownership
The full cost of running an AI system across its life: model and inference spend plus integration, human review, exceptions, recovery, governance and maintenance — not just the price per call.
Cost-migration test
A check that traces every apparent AI saving across budget owners to see whether the cost was removed or merely moved into review, exceptions, recovery, governance or adoption.
Dependable outcome
A result from an AI-assisted workflow that is actually usable without further repair — the right denominator for measuring the human attention and cost AI automation really consumes.
Used in these essays
The wrong questions about AI right now
Many of the questions that helped us orient ourselves around generative AI are now too blunt to be useful. The harder work is no longer asking what AI is in the abstract, but specifying where it works, where it fails, what authority it should have, and what the whole system costs.
A cheaper AI model can move the cost instead of removing it
A lower model bill can hide a higher workflow bill. The useful AI TCO question is not only what got cheaper, but where the cost moved.