Operators’ Corner · v2
AI token economics: from model usage to planned business value
By Richard White ·
Free, no form. The workbook uses illustrative values; replace them with your own.
Abstract
A product-economics supplement compares supported reuse with separate delivery, counting consumer onboarding, compatibility, support and full operating cost without treating avoided effort as realized cash savings.
AI spending is easier to approve than to justify. This guide gives finance, technology and business leaders a shared economic model for API-based assistants and agents: forecast demand and cost by workflow, measure cost per accepted task, and release funding against evidence of quality and value.
It separates capacity released from cash realized, shows fully allocated, avoidable and incremental cost side by side, tests a business downside and rising prices, and tests whether negotiated prices and growing usage could change the economics. A companion Excel workbook implements the method with illustrative values you replace with your own.
Decision it supports
Whether the workflow’s economics justify expanding, holding or stopping.
What you will produce
A cost per accepted task, a capacity-to-cash bridge and an expansion decision.
See it worked through in a guide
Claims package 1.1.0 · 26 October 2026. Synthetic worked example: team A, group disability, region 2. Read-only preparation of the assigned case; examiners retain every case decision. Hold candidate 1.0.0 after an unrelated employee’s claim is returned. In the synthetic decision, accept 1.1.0 for a monitored team A rollout only after prior-claims history is disabled and the remaining scope passes a fresh evaluation. All figures, evidence and sign-offs in that guide are invented; the paper provides the underlying method.
In the paper
Three passages, as published
Approve versus justify
AI spending is easier to approve than to justify.
Capacity is not cash
Finance can challenge and reconcile the assumptions. It can’t make a department redeploy capacity.
Hurdles
Falling below a hurdle doesn’t prove that stopping helps. Clearing it isn’t enough authorization to scale.
Next step
Discuss this paper against your own program.
A free sixty-minute working session with Richard on your top gaps. You leave with a one-page summary within three business days, and three next questions.