OpenAI’s 69-page report on ChatGPT enterprise adoption doesn’t address the biggest question: ROI

Welcome to Eye on AI.

Emily Forlini here, filling in for Jeremy one last time as his vacation comes to a close.

In today’s issue: Juicy details OpenAI doesn’t want you to see in its new report Anthropic reportedly plans a $2 trillion IPO in October—the largest ever OpenAI replaces its chief revenue officer after less than a year Google pronounces Sam Altman dead—for 41 minutes It really sunk in for me this week just how much money is flowing in the AI industry.

I spoke with two former OpenAI employees who made about $10 million in a day by selling shares in an internal tender offer, which Bloomberg reports totaled $7 billion across the staff.

Then, this morning, on Fortune ‘s weekly AI podcast, my coworker Beatrice Nolan and I interviewed the CEO of Lovable.

This week, the old Stockholm-based firm, which is only three years old, doubled its valuation to $13.3 billion.

A couple million, a hundred billion, a trillion (or two, in the case of Anthropic’s upcoming IPO )—what’s the difference at this point? There’s just one big problem looming in the background: The ROI of AI adoption is still not clear for companies.

OpenAI grapples with this existential question in a 69-page report published on August 11 on the enterprise adoption of ChatGPT.

On its face, the report tells the a story of exponential AI usage growth across all seniority levels and job functions, highlighting what it calls a “frontier gap,” in which companies who are using AI are pulling ahead of those who aren’t.

In other words, if you’re not using AI—especially agents you can delegate tasks to—you’re losing.

But the fine print tells a different story.

We still don’t know if AI helps you make more money In one small table on page 35, the researchers report no statistically significant correlation between the revenue per employee, and how much those employees use AI, measured in messages sent and tokens used.

“Revenue per employee is not meaningfully associated with output tokens per employee or messages per active user once other controls are included,” the report explains.

Importantly, the revenue numbers here are from before the employees began using ChatGPT.