A scorecard for the AI age

OpenAI published a framework for measuring the value of artificial intelligence investments on July 17, 2026, arguing that businesses should evaluate AI based on work accomplished rather than adoption metrics such as seats purchased or licenses renewed. The company said the basic economic question for leaders is whether the value of work AI completes grows faster than the cost of producing it. OpenAI proposed a scorecard centered on "Useful Intelligence per Dollar" to answer four questions: how much useful work gets done, what a successful task costs, how often AI gets the work right, and whether each AI dollar produces more value as usage grows. The company said the lowest price per token does not always produce the lowest cost per outcome because a more capable model may complete a task in one pass while a cheaper model requires retries and human review. OpenAI cited its GPT-5.6 family, released last week, as an example of a tiered model lineup designed to optimize cost per successful task. The company said GPT-5.6 Sol set a new state of the art on the Artificial Analysis Coding Agent Index while using 54% fewer output tokens than another leading model. OpenAI said dependability has direct economic value because accurate results reduce time spent reviewing and correcting work, and that organizations should define boundaries for data access and system actions before AI moves from drafting to taking action.