Uber president and chief operating officer Andrew Macdonald says the company’s fast‑growing AI costs are getting harder to justify, largely because he can’t identify a clear link between AI use and real productivity gains.
In a new interview with the YouTube channels Rapid Response and Masters of Scale, Macdonald points out that while some internal numbers show more code is being written using Anthropic’s Claude Code—the company’s top AI budget item—it’s unclear whether that code is making a difference on either side of the Uber experience. Having blown the company’s entire AI budget on Claude Code in just the first four months, Macdonald now argues that spending heavily on AI tools rather than hiring human staff only makes sense if the company can prove these systems help it ship better products more quickly.
Uber has leaned hard into AI across its business, from pricing algorithms to internal copilots for engineers, legal teams, and marketers. CEO Dara Khosrowshahi recently said that AI agents contribute around 10% of Uber’s code changes, with humans still reviewing everything. Even this could get trickier to justify, however, as more businesses grow dubious about the claim that generative AI can save (or make) a lot of money.
