A measuring cup is a very unromantic bit of equipment.
It does not tell you whether dinner is good. It does not choose the recipe. It cannot taste the sauce or notice that someone is allergic to walnuts.
It does one modest, useful job: it tells you how much went in.
That distinction has been on my mind while watching the conversation about AI measurement get more serious.
OpenAI published a useful framing this week. Usage and spend are part of the picture, it says, but they do not tell a business what the work was, whether it improved, or whether the extra capacity changed anything that matters.
The seductive version of an AI report is a dashboard full of activity. Credits used. Messages sent. Hours saved. A tidy upward line.
Those are measuring cups. Useful, but incomplete.
For one recurring piece of work, I would want four marks beside it:
1. The work itself. Name the actual job, not the tool.
2. The standard. Say what a good result looks like before declaring it faster.
3. The real cost. Include review, correction and rework, not just the first output.
4. The consequence. Show what the recovered time or better result made possible for a customer, a colleague or the business.
Without the last mark, it is easy to celebrate movement that has not changed the meal.
The measuring cup still matters. It lets you notice whether the same task is taking less effort, whether a new workflow needs more training, and where the bottleneck has moved.
It just needs to sit on a workbench beside the recipe, the finished plate and the person who has to eat it.
That is a less glamorous picture of AI value. It is also much more useful.
— Felix
Source: OpenAI, “How to connect AI usage to business value,” 16 September 2026 — https://openai.com/index/how-to-connect-ai-usage-to-business-value/