Uber's Chief Operating Officer: Returns from AI investments harder to justify, token consumption hasn’t brought more useful features

Uber Chief Operating Officer Andrew Macdonald stated in the Rapid Response interview released last Saturday that it is increasingly difficult internally to justify AI expenditures. Previously, Praveen Neppalli Naga, Uber's technology chief, mentioned in an interview with The Information in April that the company has already exhausted its 2026 Claude Code budget.

Macdonald said that this statement has sparked internal discussions about AI token consumption and has also led the management to re-examine the trade-offs brought about by AI investments, including staff allocation. Uber CEO Dara Khosrowshahi earlier this month stated during an earnings conference call that the company is slowing down its recruitment to offset AI investments.

Macdonald indicated that after communicating with Uber's senior engineering leadership, he found that despite a significant increase in token usage, it has not yet translated into a proportional increase in useful consumer functions.

“The correlation isn’t there yet. Intuitively, it seems like more things are being produced, but it’s hard to directly map the token usage data into ‘we are truly producing 25% more useful consumer features’.”

He added that as an ordinary user trying various AI use cases for free, AI easily gives the illusion of being “free”; but in a corporate environment, all usage will eventually translate into real and escalating costs.

Currently, attitudes towards AI within major tech companies are diverging. Some companies are still actively encouraging the frequent use of AI, even including it in employee performance assessments; while others are beginning to adjust their direction. Duolingo once reversed its decision to include AI usage in performance assessments, as employees questioned whether such practice would lead to formalism “for the sake of using AI.”