AI Unit Economics Calculator

The cost and margin of AI work measured in business units: cost per completed task, margin per task, AI cost per customer. It differs from traditional software economics because the marginal cost of serving one more unit is not near zero; every task consumes paid compute, and failed attempts consume it without producing anything.

Divide total spend over a period by the number of tasks that finished successfully in that period. Failed attempts stay in the numerator, which is the point: three billed attempts that produce one good answer give a cost per completed task of three attempts, not one. The calculator adds retries and human review time on failures for you.

There is no universal bar, but a useful reading: cost well below revenue per task is healthy, cost close to revenue is an engineering problem with known levers, and cost above revenue at any plausible efficiency is a pricing or product problem that optimisation cannot fix. The verdict card applies exactly that reading.

Cost per completed task multiplied by the tasks a customer consumes per month. It is the number to put next to your subscription price, and it is the difference between a customer who is profitable and one who quietly is not. Heavy users on flat plans are where AI margins go to die.

Because it improves when quality gets worse. Failures and retries make each call look cheap while making each finished task expensive, so a per-call dashboard rewards exactly the regression you most need to catch. Cost per completed task moves in the honest direction, rising when retries rise.