OpenAI Report Finds No Link Between AI Use and Revenue Per Employee

AI use and revenue per employee: OpenAI’s surprising findings

OpenAI’s freshly released 69‑page enterprise study reveals that AI use and revenue per employee are not statistically linked, challenging the hype surrounding AI‑driven profit spikes. The report, posted on Aug 11, shows a steep climb in ChatGPT adoption across job tiers, yet a tiny table on page 35 reports no meaningful correlation between employees’ token volume and revenue per head.

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

While high‑earning firms tend to be early adopters, the data stop short of proving that heavier AI consumption translates into higher earnings. Executives appear the least enthusiastic users; a chart on page 29 shows senior leaders generate the fewest weekly messages, whereas junior staff dominate usage. OpenAI CFO Sarah Friar warned on LinkedIn that competitive advantage lies with the workers closest to the front line.

Enterprise usage plateaued from Oct 2025 to Dec 2025, a period when Anthropic’s Claude Code surged. By Jan 2026 the trend rebounded, with OpenAI reporting a ↑ 27% lift in token output as existing clients deepened engagement and new accounts signed on. The company has also reshuffled its sales leadership, installing Dali Rajic as chief revenue officer after Denise Dresser’s brief tenure.

According to Reuters and Bloomberg, the AI market continues to attract massive capital, but the study’s credibility is mixed: two of the five authors are paid contractors from Columbia Business School and Wharton, noted in a footnote. Their involvement blurs the line between independent research and corporate PR.

For firms still weighing AI bets, the takeaway is clear: monitor on‑the‑ground impact rather than rely on glossy forecasts. The pandemic context taught that disruption can be swift; today’s AI rollout may follow a similar unpredictable path.