Remember the high-profile story about tokenmaxxing, when Uber's CTO admitted that the company had already completely "burned" its budget for Anthropic's Claude Code, which was designed for 2026? This time, the top manager came back with a completely different story, or rather, with experience in implementing AI.
Recall that tokenmaxxing is a trend towards maximizing the use of artificial intelligence tokens that swept Silicon Valley this spring. However, having received cosmic checks, IT giants began to curtail such generosity. In particular, at Uber, employees managed to spend their entire annual AI budget in just four months. To keep costs under control, they set a clear ceiling — a maximum of $1,500 per month per developer for coding assistants like Claude Code or Cursor.
After this limitation, Uber came up with the idea of sending its best AI engineers to departments like finance, legal, marketing, customer support, HR, and procurement. The goal is to learn how people actually work before developing AI tools for them, Business Insider reports .
Uber CTO Pravain Nappalli Naga called these two-week teams “Agentic Pods.” First, engineers spend a few days observing employees, studying every step of their workflow, and then quickly develop and test software to improve it.
The results are impressive. One financial planning process has been reduced from 15 hours to 30 minutes. Financial reporting has been reduced from two days to 10 minutes. And marketing quality assurance has been reduced from two weeks to less than an hour.
Naga noted that the biggest results rarely come from simply speeding up a single task. They come when the entire workflow is completely reengineered for AI: eliminating unnecessary approvals, replacing outdated software, and helping people make decisions faster. This is clearly evident in this case.
A number of top IT managers have compared this approach to the “forward-deployed engineer” role model, which is rapidly gaining popularity in Silicon Valley.
Recall that the US is experiencing a real personnel boom around forward-deployed engineers : the number of vacancies for these specialists has increased by more than 700% in a year. In addition, Amazon is investing $1 billion in creating a separate new team of such AI engineers .
But Uber has taken a slightly different path: instead of sending AI experts to external clients, the company is involving them within its own business, looking for new opportunities to increase productivity.
