OpenAI has added usage analytics and spend controls to ChatGPT Enterprise, letting admins set a default credit limit across an entire workspace, override it per group, and grant exceptions to individual employees. The console now breaks credit consumption down by user, product and model, spanning both ChatGPT and Codex.
The framing is the tell. OpenAI notes that seat-based pricing is easy to model on day one and hard to model on day 400, once staff move from occasional lookups to embedded workflows that hit the model dozens of times an hour. Employees can watch their own credit burn and request more, turning consumption into something a finance team can actually govern.
This is the bill arriving for agentic work. As one chat turns into a loop that calls the model repeatedly, cost stops tracking headcount and starts tracking behaviour, and the people who feel it first are the admins now staring at a per-user credit breakdown.