OpenAI Studies Its Own ROI. Finds Nothing. The Timing Is Suspicious, Obviously.
OpenAI's research examined whether ChatGPT corporate customers see a clear return on investment. The study found no correlation between AI use and revenue per employee. However, the revenue figures were collected before employees began using ChatGPT, not after. Fortune's Eye on AI newsletter flagged this rather significant methodological detail.
This is a textbook case of what researchers call temporal mismatch. The mechanism: you cannot measure the effect of an intervention using data from before the intervention occurred. It would be like weighing yourself before a diet and declaring the diet ineffective. The broader lesson for anyone evaluating AI tools is that ROI measurement requires before-and-after data on the same metric. Without that, you are reading tea leaves.
OpenAI conducted the research on its own corporate customers. Fortune's biweekly Eye on AI newsletter reported and clarified the timing discrepancy.
- Pick one recurring task in your workweek, such as drafting emails or summarizing documents, and note how long it takes you without AI.
- Use ChatGPT or any free AI assistant for that same task over one week. Track the time spent including any editing.
- Compare the before and after times. That delta is your personal productivity data. Without both numbers, you have no ROI story, only an anecdote.