OpenAI's CFO says forget model benchmarks—here are the 4 metrics that actually measure AI ROI

openai cfo sarah friar introduced a new 'ai scorecard' designed to measure the return on investment of ai deployments through four specific metrics: useful work, cost per successful task, dependability, and return on compute. by focusing on these operational and financial metrics rather than traditional model benchmarks, the scorecard addresses the gap between technical capability and business value. this move targets enterprise customers who need concrete data to justify escalating ai infrastructure and usage costs, shifting the industry conversation from what models can do to what they actually deliver in production.