Uber CTO Praveen Neppalli Naga who sparked Tokenmaxxing panic has another lesson on using AI

Earlier this year, Uber’s CTO Praveen Neppalli Naga triggered a “tokenmaxxing panic” after revealing the company nearly blew through its 2026 budget for Anthropic’s Claude Code in just three months. Now, Naga is back with a new lesson. According to a report by Business Insider, Naga said that the biggest gains from AI don’t come from speeding up single tasks, but from redesigning the complete workflows. Naga said that Uber has started deploying its best AI engineers into departments like finance, legal, marketing, HR and procurement. These two-week teams, dubbed Agentic Pods, shadow employees for several days to understand workflows, then rapidly build and test AI-powered tools.The results have been dramatic:* Financial planning cut from 15 hours to 30 minutes.* Financial reports reduced from two days to 10 minutes.* Marketing quality checks shrank from two weeks to under an hour.

Uber CTO says productivity boost comes from workflow redesign

Naga emphasised that the real productivity boost comes from workflow redesign — eliminating redundant approvals, replacing outdated software, and enabling faster decision-making. Several executives compared the approach to Silicon Valley’s forward-deployed engineer model, but with a twist: instead of embedding engineers with customers, Uber is embedding them inside its own operations.Vantortech CPO Peter Wilczynski jokingly dubbed the role the “Rearward Deployed Engineer”, describing specialists embedded deep inside corporate processes to reimagine how work gets done with AI.

Uber CTO says company is moving beyond ‘Tokenmaxxing’

To bridge the gap between high AI costs and real-world efficiency, Uber launched a project called “Agentic Pods”. The goal was to stop treating AI as an expensive experiment and start using it to redesign complex, manual workflows across departments like Finance, Legal and Marketing.Naga said that the company handpicked approximately 30 of its most AI-proficient engineers and paired them directly with domain experts – the employees who actually perform the work. The teams operated on a strict, two-week schedule. The claimed that in just two months, Uber successfully ran 16 of these “Agentic Pods” across 16 different business functions, and noted that the efficiency gains were significant.