News
Uber Embeds AI Agents in Finance and HR, Cutting Reports From Two Days to 10 Minutes
Uber is moving its top AI engineers into finance, HR and legal departments to build so-called agentic pods. The result: financial reports that once took two days now take 10 minutes, and capital allocation across 150 cities dropped from 15 hours to half an hour.
Contents
Uber launched a program in which the company's most AI-experienced engineers spend two weeks embedded in non-engineering departments, such as finance, HR, or legal, working alongside employees there to build AI agents tailored to specific, everyday tasks. The company calls these teams "agentic pods," and Uber's CTO, Praveen Neppalli Naga, says the model delivers results that simply buying AI tool licenses could not achieve.
How an Agentic Pod Works
Rather than outsourcing process automation to outside IT teams, Uber sends engineers directly to the people who run the process day to day. Over two weeks, an engineer observes a finance, HR, or legal team at work, talks to employees about where they lose the most time, and builds an AI agent on the spot that takes over the repetitive parts of that work.
Naga explains that process documentation rarely reflects how work actually happens in practice. Direct observation, he says, gives agents a real shot at taking over tasks that look different on paper than they do when teams actually perform them.
You can't automate these effectively by looking at process diagrams or documentation. You have to understand how the work is actually done. - Praveen Neppalli Naga, Uber's Chief Technology Officer
Numbers That Stand Out
Uber points to two flagship results from the program. The financial pacing report, previously prepared over two days, now takes 10 minutes. Capital allocation across the 150 cities where the company operates, a task that used to consume 15 hours of analysts' work, now takes 30 minutes.
The company ran 16 such agentic pods over two months and says the model will be expanded. Naga talks about building a dedicated team to deepen this way of working and redesign further processes from the ground up, rather than simply layering AI on top of existing workflows.
We're now building a dedicated team to scale this and go deeper. They'll understand the work deeply, redesign it from the ground up, and use AI to fundamentally change how the business operates. - Praveen Neppalli Naga, Uber's Chief Technology Officer
Tension Over AI Bills
The agentic pods program arrives as Uber itself grapples with rising AI tool costs. In spring 2026, the company exhausted the budget set aside for Claude Code, forcing cuts to further spending on AI-powered development tools. Uber COO Andrew Macdonald acknowledged in May that justifying further AI investment is getting harder, since large outlays don't always translate into features riders or drivers can actually see.
Agentic pods can be read as a response to that pressure: instead of continuing to increase spending on licenses, Uber is trying to squeeze the most out of the resources it already has, directing its best engineers toward places where the payoff can be measured in hours and days of administrative work saved.
What It Means for Companies in Poland
Uber's model shows a different approach to rolling out AI across a large organization than the chatbot or coding-assistant license purchases common in Poland for entire teams. Instead of handing out tools and hoping employees pick them up on their own, the company invests the time of its best AI specialists into specific, narrow processes in departments not normally associated with technology.
For Polish companies just starting to think about AI agents in finance or HR, this is a signal that the scale of savings can be real, but it requires the involvement of people who know the process from the inside, not just a subscription purchase. Analyst reports have previously noted that Polish companies are adopting AI slowly mainly due to skills gaps, not lack of access to the technology, which could make the agentic pods model difficult to replicate without the right internal resources.
Uber has not disclosed how much it cost to launch the first 16 agentic pods, nor how it calculates the return on that investment against its previously constrained AI budget. The company says it plans to keep scaling the program, meaning other departments, likely customer service or logistics, could go through a similar process in the coming months.
Sources: How Uber uses AI for development (newsletter.pragmaticengineer.com), Uber's CTO embedded its top AI engineers in HR, finance, and legal, and found better ways to build (aol.com)
