Uber has pushed its gig-work machine into AI data labeling, but firing the two executives who built that unit makes the bet look less settled than the company says.
Uber just removed Naga Kasu and Pankaj Kamat from the AI data-labeling business they helped build, and the company isn't offering much beyond the phrase every reader should treat carefully: leadership transition. According to Bloomberg, Kasu, a senior director of engineering, and Kamat, a director of product at Uber AI Solutions, were let go this week after more than a decade each at the company. An Uber spokesperson confirmed the exits to Bloomberg, said the division is seeing strong momentum, and didn't name replacements.
You don't dismiss the two people who built something from scratch and expect the word momentum to answer the obvious question. What actually changed inside the unit?
Uber AI Solutions began in November 2024 as Scaled Solutions, TechCrunch reported at the time. The idea was plain enough: take the operating system Uber built for matching people to rides, food deliveries and time-sensitive tasks, then sell that labor network to AI companies that need humans to check model outputs. Workers validate code, review audio in different languages and label video used in autonomous-vehicle systems. Forbes has reported that Uber counts Alphabet, Aurora Innovation and Niantic among its customers, and that the service has expanded to more than 30 countries after launching in five.
That makes the exits more than a personnel note. Kasu and Kamat weren't celebrity AI hires brought in for a press release. They were Uber operators, people who already understood the marketplace logic behind dispatch, routing and quality scores before applying it to data work. If you're trying to convince enterprise AI buyers that Uber can handle the dull, exacting labor behind model training, that kind of institutional knowledge matters.
The market Uber entered got noisier after Meta's Scale AI deal in June 2025. Reuters and other outlets reported that Meta agreed to invest about $14.3 billion in Scale AI, with founder Alexandr Wang moving into a senior role at Meta's superintelligence effort. That deal made customers and rivals reassess their dependence on Scale. Companies including Mercor, Turing, Invisible Technologies and Surge AI have been fighting for the work left in the scramble. Surge, founded by Edwin Chen, has been reported by The Information and Reuters to be running at a scale that would have sounded absurd for a bootstrapped data company a few years ago.
Uber's pitch is different from the startup pitch. It doesn't need to persuade investors that it can manage a distributed workforce, because it already does that every day in ride-hailing and delivery. Its 2025 annual revenue was about $52 billion, according to its public financial reporting, and the company has more operating machinery than most AI-labeling startups will ever build. Frankly, that is the best argument for the whole project. If AI labs need armies of people to grade, correct and sort model behavior, Uber already knows how to move armies of contractors through an app.
But data labeling isn't just another delivery slot. A late burrito and a bad reasoning trace are different failures. Enterprise AI customers care about consistency, domain knowledge and repeatable evaluation standards, not only whether a worker accepted a task and finished it. Uber has reportedly explored using idle drivers for annotation work during downtime, which sounds efficient until you ask what kind of work is being routed to whom, how the quality is checked, and whether a driver waiting between trips is the right person to judge a code answer or a medical transcript.
That is the gap Uber now has to close without the two executives most closely tied to the product and engineering buildout. The company may have good internal reasons for the firings. It hasn't given them publicly. What readers can see is narrower: a young unit in a crowded market has lost the leaders who knew both sides of the bet, Uber's marketplace plumbing and the AI industry's appetite for human judgment.
Uber also hasn't disclosed revenue for AI Solutions, so there's no public number that proves the business is working at meaningful scale. Customer names and country counts show reach, but revenue, retention and quality metrics would tell you far more. Until Uber shares those, the safest reading is that the unit remains a serious experiment rather than a proven new pillar.
The point isn't that Uber's AI labeling push is finished. It isn't. The point is that the hard part starts after the clever platform analogy. Anyone can say gig work can be repointed at AI. Uber now has to show that it can keep enterprise customers, protect quality and replace experienced internal leaders without losing the operating discipline that made the bet plausible in the first place.
Also read: Together AI Raises $800 Million From Aramco at an $8.3 Billion Valuation • Venice AI became a unicorn by promising to forget everything you tell it • Cisco is betting its whole workforce on AI agents while it cuts jobs
