First hard data on how organizations actually use enterprise ChatGPT

Most of what we hear about AI adoption inside companies is guesswork dressed up as insight. This study is different. Researchers from OpenAI, Wharton, and NBER got access to actual ChatGPT Enterprise account records and matched them to worker roles, task types, and financial data from public companies, through March 2026. That’s over 1,500 organizations and more than 17 million messages, tracked to the six month mark of adoption. This is one of the first real looks at what’s actually happening inside businesses using ChatGPT, not what vendors claim or what surveys guess.

A few patterns stand out. Adoption is concentrated in larger companies, especially ones that already spend heavily on R&D and SG&A. That tracks with what most owners already sense, bigger firms with more slack in the budget move first. The more interesting finding is about who inside these companies actually uses the tool the most. It’s not the senior staff. It’s early-career workers who show the highest usage intensity. If you’ve been on the fence about whether junior hires can get real value from AI tools versus veterans who “already know how to do the job,” this is a data point worth sitting with. It suggests the payoff may show up faster with your newer people than with your most experienced ones.

For a business owner, the takeaway isn’t “buy more seats.” It’s that adoption patterns are not random, and if you’re rolling out AI tools without watching who uses them and how, you’re missing signal that’s now backed by real usage data. Tracking that kind of thing consistently, who’s using what and where the value shows up, is exactly the kind of thing we build into an AI command centre, Omni. You can read the full research paper here if you want the details.