TLDR: Uber COO Andrew Macdonald says AI spending lacks a direct link to productivity gains, complicating headcount cuts.
Key Takeaways:
- Uber embedded AI across engineering and used Claude Code for about 25% of code commits last quarter.
- Andrew Macdonald said the link between productivity gains and shipped user value is still missing.
- If AI returns stay murky, firms will keep scaling spend harder while slowing hiring or cutting roles to pay.
Uber is seeing the same thing many AI believers do not want to admit: code churn and flashy agents do not automatically become customer value. Until the bill shows up in results, leadership will treat hiring like a safety harness, not a growth lever.
Uber is seeing the same thing many AI believers do not want to admit: code churn and flashy agents do not automatically become customer value. Until the bill shows up in results, leadership will treat hiring like a safety harness, not a growth lever.
Q&A
What will Uber need to prove to reconnect AI usage with user outcomes?
Executives will likely demand measurable links between AI based development workflows and specific shipped features, retention metrics, or cost per successfully delivered project.
Why would AI token costs push companies toward internal incentives rather than budgets?
If tokens come from company accounts, incentives can shift to employees for driving usage, masking spend by framing it as performance, not consumption.
How could agentic commerce change Uber's priorities if it finally takes off?
Uber may shift from treating automation as a backend efficiency tool to redesigning user acquisition, marketplaces, and partnerships to compete with assistants that transact directly.
What does Nvidia CEO Jensen Huang's token framing imply for software teams under budget pressure?
Teams may face higher expectations to justify AI consumption, but those expectations collide with the reality that token spend must translate into shipping output, not just experimentation.
If productivity gains stay hard to justify, what second order effect could hit the wider U.S. AI buildout?
Funding could move from broad AI infrastructure expansion to narrow, high ROI deployments, slowing scaling plans across industries that invested ahead of measurable payoff.
No comments yet. Be the first to share your thoughts!