4OpenAI Blog·
OpenAI CFO introduces practical AI ROI scorecard framework
Sarah Friar, CFO of OpenAI, published a framework for measuring AI return on investment using four metrics: useful work, cost per successful task, dependability, and return on compute. The piece is positioned as practical guidance for organizations evaluating AI deployments. It signals OpenAI's interest in shaping enterprise adoption narratives around measurable business outcomes.
Related guides (3)
Related events (8)
OpenAI publishes enterprise guidance on managing AI investments in the agentic era
OpenAI published a blog post advising enterprises on how to manage AI investments as agentic workflows become central, framing the key metric as 'useful work per dollar' rather than traditional cost measures. The post covers efficiency improvement and scaling of high-value agentic workflows. While light on technical substance, it signals OpenAI's positioning toward enterprise buyers evaluating agentic ROI.
The State of Enterprise AI 2025 Report
OpenAI has published a report summarizing key findings from its enterprise customer data, highlighting accelerating AI adoption, deeper integration into workflows, and measurable productivity gains across industries in 2025. The report draws on OpenAI's own enterprise deployment data rather than third-party surveys. It serves as both a market signal and a strategic communication about the business traction of OpenAI's enterprise offerings.
OpenAI launches Economic Research Exchange to study AI's labor and productivity impacts
OpenAI has announced the Economic Research Exchange, a program to fund and facilitate external research on AI's effects on jobs, productivity, and the broader economy. Applications are open for selected research projects. The initiative signals OpenAI's interest in shaping the empirical narrative around AI's economic consequences.
OpenAI publishes framework modeling AI impact across 921 U.S. occupations
OpenAI released an analytical framework examining 921 occupations and 148 million U.S. jobs to categorize roles by automation risk, reorganization potential, growth, or minimal AI disruption. The work represents OpenAI's own modeling of labor market transitions driven by AI. It is notable as a primary-source policy-adjacent analysis from a frontier lab, signaling OpenAI's engagement with workforce displacement questions.
OpenAI Introduces GDPval: Evaluation of Model Performance on Economically Valuable Real-World Tasks
OpenAI has released GDPval, a new benchmark designed to measure AI model performance on real-world economically valuable tasks spanning 44 occupations. The evaluation aims to move beyond traditional academic benchmarks by grounding model assessment in tasks with direct economic relevance. This represents OpenAI's effort to better quantify the practical utility and labor-market impact of frontier models.
OpenAI Recapitalization Announced with Mission-Focused Governance Framing
OpenAI has announced a recapitalization intended to strengthen its mission-focused governance structure. The move is framed as expanding resources to ensure AI development benefits everyone while advancing innovation responsibly. This represents a structural corporate change at one of the most influential AI labs, with implications for how OpenAI balances commercial and nonprofit objectives.
Making AI Work: Leadership, Lab, and Crowd
This commentary from One Useful Thing proposes a framework for organizational AI adoption centered on three elements: leadership commitment, structured experimentation (lab), and distributed employee engagement (crowd). The piece offers practical guidance for companies navigating AI integration. As a tier-2 commentary source, it reflects practitioner thinking on enterprise AI deployment patterns rather than reporting new technical developments.
Measuring AI's capability to accelerate biological research
OpenAI introduces a real-world evaluation framework designed to measure how AI systems can accelerate biological research in wet lab settings. The work uses GPT-5 to optimize a molecular cloning protocol as a concrete demonstration case. The framework explicitly addresses both the potential benefits and biosecurity risks of AI-assisted experimentation, positioning this as a dual-use capability assessment.