Why Big Companies Are Suddenly Cracking Down on AI Spending

Many businesses are slamming the brakes on employee AI use after realizing the hidden costs of these tools. Instead of a small monthly fee, companies now face bills that explode based on how much workers use AI—like paying for electricity by the kilowatt instead of a flat rate. Suddenly, departments that once embraced AI assistants are seeing their budgets drain in days. Some firms, like Atlassian, saw monthly AI expenses jump from five million to over fifteen million dollars in less than a year. Others, like Citi, blocked access to the latest AI models entirely, forcing employees to downgrade to weaker versions that cost less per use.

The problem started when companies jumped into AI without fully grasping how the billing works. Most providers now charge by the "token"—tiny pieces of data processed during each interaction. As a result, frequent users can burn through company resources quickly, especially when teams run complex AI agents all day. One employee complained that colleagues rebuilt their daily workflows around AI, only to hit budget walls within hours. Others panicked when their tools got locked, asking in group chats, "Now what do I do? " Some industries, such as finance and software, were hit hardest because their work relies heavily on AI-assisted tasks. Financial firms like Citi warned staff to avoid the flashiest models for simple questions, suggesting cheaper alternatives for coding or text generation. Still, even those measures weren’t enough to prevent overspending. Reports show other companies faced even wilder surprises—like a single month’s bill reaching half a billion dollars for one firm or an unsuspecting developer getting charged eighteen thousand dollars for a minor security test.

Actions