Uber Technologies has placed usage limits on some artificial intelligence-powered tools used by its employees. The move is aimed at controlling costs after the company ran out of its artificial intelligence budget earlier this year.
Ride-hailing giant Uber is capping token spending per AI coding tool for all employees at $1,500 per month, an Uber spokesperson told Bloomberg News. This means spending on one tool does not affect the budget for another tool. This limit was introduced in recent months and only applies to agent coding software such as Cursor and Anthropic PBC’s Claude Code.
Every employee has a dashboard where they can track their usage of various tools. The company has also introduced a process that allows individuals to request permission to exceed the normal cap.
“We believe this is all a very easy way to responsibly accelerate the adoption and experimentation of agent AI across the enterprise,” the spokesperson said.
This restriction, which has not been reported, was put in place in response to Uber’s increasing adoption of AI tools within the company. Chief Technology Officer Praveen Nepali Naga told the Information in April that the company had already maxed out its AI budget for the year. CEO Dara Khosrowshahi said last month that about 10% of the company’s code is submitted and built by AI agents, with increased usage by legal and marketing teams.
In addition to Uber, companies in a variety of sectors are seeking to improve productivity through AI tools while also trying to control related costs. Uber announced last month that it would be slowing down its overall hiring pace compared to this year’s plan to capitalize on the benefits of leveraging AI within the company.
Still, it remains to be seen whether the focus on AI tools will lead to more new features for Uber customers, according to Chief Operating Officer Andrew McDonald. “It’s very difficult to draw the line between one of those statistics and, ‘Okay, now we actually have like 25% more useful features for consumers,'” he said on the Rapid Response podcast last month.
“It may become clearer in the coming quarters and years, but I think it’s difficult today, even if some of the underlying metrics are going in a really astronomical direction,” he said on the show.
Mr. Lunn writes for Bloomberg.
