Uber is setting limits on how much employees can spend on artificial intelligence agents after its workers blew through their yearly AI coding budget in just four months.
Each employee at the rideshare and food delivery company is limited to $1,500 in spending on tokens per AI coding agent, according to Bloomberg. The limits apply to only AI software used to write and execute code, as opposed to more general AI agents.
AI tokens are small units of text and data that AI models read, remember and generate, explains Microsoft.
In April, Uber Chief Technology Officer Praveen Neppalli Naga told technology industry news site The Information, “I’m back to the drawing board because the budget I thought I would need is blown away already.”
Uber CEO Dara Khosrowshahi attends the …
more >
Mr. Neppalli Naga said the main AI coding software used at Uber since December was Claude Code, developed by Anthropic, followed by competing AI coding software Cursor. He also said Uber plans to let its engineers try out OpenAI’s competing coding agent Codex.
In a May earnings call, Uber CEO Dara Khosrowshahi said that around 10% of Uber’s code is developed by AI agents, according to Fortune.
Some at Uber are wondering whether AI use is effective yet.
Uber Chief Operating Officer Andrew Macdonald told the Rapid Response podcast last month that even with AI usage going up, “maybe implicitly there’s more that is getting shipped, but it’s very hard to draw a line between one of those stats and ‘OK, now we’re actually producing like 25% more useful consumer features.’”
Uber is not the only company trying to rein in employee use of AI. Sources familiar with the matter told Bloomberg that Walmart has also capped how much its employees can use the company’s in-house AI agent.
