A scorecard for the AI age

OpenAI proposes a framework for measuring AI value called "Useful Intelligence per Dollar," which focuses on work accomplished rather than traditional software metrics like user adoption. The framework asks four key questions: whether AI completes meaningful work, what each successful task costs, how reliably the system produces correct results, and whether AI becomes more cost-effective as usage scales. This matters because business leaders need to understand whether the value created by AI work grows faster than the cost of producing it, since cheaper models may require more attempts or human review while more expensive models might complete tasks correctly in one attempt. The approach emphasizes measuring full task costs including employee time and human review, tracking whether results are ready to use or need correction, and defining clear boundaries around data access and system permissions before deploying AI to handle important workflows.