Uber's AI Agent Spending Spiral Exposes Engineering Org Scaling Problem
Source: Nate’s Substack
Uber's 95% AI tool adoption rate and agent-generated 1,800 code changes weekly exposed a gap between the automation capabilities vendors are selling and the organizational capacity companies actually have to manage, validate, and govern agent-generated outputs at scale. The company's infrastructure and oversight systems couldn't scale with the velocity of autonomous agents, forcing early budget exhaustion. For growth-stage organizations, this suggests that AI adoption ROI isn't about tool penetration rates but about the unglamorous, expensive work of building review pipelines, quality gates, and ops teams to handle what agents actually produce.