Facing pressure from CFOs to optimize technology budgets, Sarah Friar, CFO of OpenAI, has proposed a new approach to evaluate the return on AI investment. Instead of measuring software success through traditional metrics like purchased seats or active users, OpenAI argues that the true measure lies in the volume of successfully completed work. This core metric, dubbed "Useful Intelligence per Dollar," directs businesses to focus on output value rather than cost per token.
Detailed Developments
In an analysis published on July 17, 2026, CFO Sarah Friar highlighted that businesses are struggling to assess the ROI of large language models. OpenAI proposes four core questions to build an AI spend scorecard: Is AI completing work that matters? What does each successful task cost? Can people depend on the result? Does each AI dollar produce more value as usage grows? This approach helps shift the mindset from using AI as a simple drafting tool to an automated workflow system capable of independent decision-making.
Technical Analysis & Technology
According to OpenAI's analysis, a model with cheaper tokens does not necessarily yield the most optimal cost for a complete task. Businesses need to calculate the full cost of completing the work divided by the number of successful tasks that meet the quality bar. To optimize this economic equation, OpenAI designed its newly released GPT-5.6 model family into three tiers: Sol (flagship reasoning model), Terra (balanced performance and cost), and Luna (fastest and most affordable). On the Artificial Analysis Coding Agent Index, GPT-5.6 Sol set a new state of the art in reasoning while using 54% fewer output tokens compared to other leading models.
Expert Opinions & Insights
Sarah Friar pointed out that AI dependability has direct economic value. When model results are accurate, consistent, and well-governed, employees spend less time reviewing, correcting, and repeating work. To make this concrete, OpenAI advises organizations to closely track three task outcomes: Ready to use, Needs correction, and Needs escalation. Only by defining clear boundaries regarding data access and human oversight can enterprises confidently integrate AI into core operational workflows.
Impact & Future Outlook
The concept of "Useful Intelligence per Dollar" reflects the maturation of the global AI market as the initial hype phase gives way to practical financial demands. In Vietnam, where businesses are actively piloting AI to optimize processes from customer service to programming, OpenAI's evaluation methodology will serve as an important reference. The combination of more efficient compute infrastructure, optimized algorithms, and solid product design like ChatGPT Work is expected to continuously drive down the cost of successful tasks in the near future.