OpenAI CFO proposes a scorecard to measure AI's real payoff
OpenAI CFO Sarah Friar lays out a practical companies can use to judge whether their AI spending is actually paying off. It rests on four measures. First, how much genuinely useful work the AI completes.
Second, the cost of getting one task done successfully, not just the price of a single query. Third, dependability, meaning how consistently the AI delivers reliable results rather than occasional lucky wins. Fourth, , or how much real value comes back for the computing power spent running the AI.
The pitch is to replace vague claims about AI value with numbers leaders can actually track.
Key points
- OpenAI CFO Sarah Friar proposes a for measuring AI's actual payoff
- Measure 1: the volume of genuinely useful work AI completes
- Measure 2: the cost of successfully finishing one task, including retries
- Measure 3: dependability, or of reliable results
- Measure 4: , value gained versus computing power spent