Sarah Friar, CFO of OpenAI, has unveiled a new AI scorecard designed to move enterprise decision-making beyond traditional metrics like accuracy or token throughput toward measurable business outcomes. The framework centers on four core measurements: useful work (quantifying actual productive output delivered by AI systems), cost per successful task (total spend divided by reliably completed jobs), dependability (system uptime and consistency), and return on compute (value generated relative to computational resources consumed). This represents a significant shift in how organizations should evaluate AI deployments, particularly as enterprises increasingly move beyond experimentation into production-scale implementations where ROI justification has become essential for continued investment.
The scorecard's real-world applicability is evident in OpenAI's customer base. Cars24, India's largest used-car marketplace, deployed OpenAI-powered voice and chat agents to handle over 1 million monthly conversation minutes, recovering 12 percent of previously lost leads while scaling agentic workflows across teams. The framework enables Cars24 to measure not just agent interactions but the conversion pipeline impact—connecting AI output directly to revenue recovery. This represents the type of 'useful work' quantification Friar emphasizes, where the scorecard helps distinguish between AI projects that generate meaningful business value versus those that simply automate tasks without ROI impact, a critical distinction as enterprise AI budgets face increased scrutiny.
The timing of this scorecard introduction reflects growing skepticism about AI's promised returns. While enterprise adoption is accelerating, many organizations struggle to demonstrate clear ROI beyond cost reduction, creating pressure on vendors like OpenAI to provide frameworks that satisfy CFO-level accountability. Friar's emphasis on dependability and compute efficiency directly addresses concerns that AI systems remain unreliable or computationally expensive for sustained business use. By publishing practical measurement methodology, OpenAI simultaneously positions itself as ROI-focused while setting industry standards that favor vendors with proven production reliability—a market positioning move that could reshape how enterprises evaluate and budget AI initiatives.