OpenAI's chief financial officer, Sarah Friar, has introduced a practical framework for assessing whether AI investments are actually paying off. The proposed "AI scorecard" centers on four metrics: useful work, cost per successful task, dependability, and return on compute. The goal is to shift evaluation away from technical benchmarks and toward operational outcomes.
The framework treats useful work as the amount of real tasks completed, while cost per successful task ties spending directly to results. Dependability measures how consistently the system performs, and return on compute connects computing resources to the value they generate. Friar argues that a common scorecard can help companies make clearer decisions about where to deploy AI.
The proposal is conceptual rather than empirical; the source does not include specific numbers or case studies. It represents OpenAI's attempt to give businesses a shared language for AI ROI, though it is not yet an industry standard.