OpenAI is proposing a new way for businesses to answer a question that has dogged the AI boom: is any of this actually paying off?
In a company blog post titled "A scorecard for the AI age," OpenAI Chief Financial Officer Sarah Friar introduces what she calls a practical AI scorecard for measuring return on investment. Rather than judging AI by hype or headcount, the framework focuses on four measures: useful work produced, the cost per successfully completed task, dependability, and return on compute.
The pitch was first reported as an exclusive by Axios, which framed it as OpenAI's CFO pitching a new way to measure AI's value and to help firms better gauge the ROI they get from AI models.
The underlying idea is that traditional business metrics don't map cleanly onto AI systems. A model that drafts a report or resolves a support ticket isn't easily captured by conventional productivity math, so Friar's scorecard tries to reframe the value question around outcomes—did the AI do useful work, how reliably, and at what cost per task and per unit of computing power.
Coming from the finance chief of the industry's most prominent company, the proposal doubles as a way to shape how customers, and possibly investors, talk about whether massive AI spending is justified.
Why it matters: as companies pour money into AI, a shared yardstick for what counts as a payoff could determine which deployments get funded—and OpenAI is angling to define that yardstick.