Pricing Enterprise AI Applications: An Economic Framework for Valuing Work and Choosing Billing Units

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

Enterprise AI applications can create value by reducing the labor required to deliver a business service.The people who use an application, the work it performs, and the units on an invoice need not coincide.This paper distinguishes these elements and shows how they relate. It estimates savings from theshare of professional work an application can support, the time it saves after review and rework, andthe proportion of released capacity that translates into lower labor cost. Those savings bound the pricefrom above. What can be charged within that bound is set by the customer’s best alternative ratherthan by the size of the savings, and a corollary states the range of value shares a vendor can feasiblypursue. The billing unit is then a separate choice, governed by what both parties can verify and bythe incentives each unit creates. An illustrative human resources (HR) and payroll case shows thecalculation. The framework explains when per-worker subscriptions are defensible, when transactionor outcome charges are more suitable, and why a value estimate alone cannot establish the profit-maximizing price. It argues that the quantity on which the value depends most, the conversion ofreleased time into lower cost, is chosen by the buyer and cannot be verified by either party, which iswhy applications of this kind are sold by subscription rather than by result. It provides a practicalmethod for analysis and teaching without assuming a universal rate of AI productivity improvement.