tech
Executive Briefing: Uber Burned Its Entire AI Budget Early. The Bill Was Trying to Tell Them Something.
The next AI budget fight will not start because employees refuse to use AI.
TL;DR
- AI costs are exceeding budgets because frontier intelligence and open models are becoming useful, leading companies to manage agentic work with outdated controls.
- Uber's experience shows that 95% of engineers use AI tools in agent-style workflows, resulting in significant code changes but unclear customer value.
- The 'token burn' is a signal that AI has transitioned from a tool to labor that needs management, and companies lack systems to manage this unseen work.
- Companies that succeed will adopt a balanced approach between excessive spending and hard capping, focusing on an intermediate solution.
- Key areas for adaptation include understanding the AI cost curve, implementing a routing rule for AI model usage, rethinking budget models, establishing an agent-first operating model, and learning to interpret AI bills correctly.