Uber has been conducting the pilot in 12 top Indian cities for about two weeks, the people quoted earlier said, speaking on the condition of anonymity.
“The most common model, especially in pilots, is to pay workers flat per-task rates. Cash remains the preferred form of incentive, as drivers view it as tangible income rather than a perk,” said one of the people cited above. Payouts typically range from ₹10– ₹50 for annotation tasks and ₹5– ₹20 for uploading photos, often calculated on a per-minute or per-click basis. These are sometimes bundled into a driver’s weekly earnings along with ride or delivery income.
Uber has about 1.4 million driver-partners in India, and the company is monetising its fleet’s idle time. Platforms with physical touchpoints—fleets, IoT devices, connected vehicles—benefit from using idle time to generate precious labelled data that can be used to train internal AI models or sold to third parties.
Global demand for high-quality labelled data is projected to reach $17 billion by 2030, according to estimates by the policy research and advocacy group Consumer Unity & Trust Society (CUTS) International. A 2021 report by Nasscom estimated that India is expected to service over $7 billion of the global annotation market by 2030.
As drivers flag lanes, intersections, or curb rules, these inputs are folded back into Uber’s training systems, which will help AI “see” the city more like a human, said the first person quoted earlier. The payoff is better routing, sharper ETAs, and fresher maps since drivers notice roadworks, new one-way streets, or shuttered storefronts long before official feeds do, the person said.
Labelling and annotation
While Uber announced in the US on 17 October that it would offer similar gig work to drivers when they are not on the road, the company confirmed to Mint that the model is already being tested in India.
The labelling extends beyond road features to include imagery and other datasets, but details remain limited. It is also unclear whether Uber is developing its own proprietary annotation tools or relying on third-party platforms.
“It’s an attempt at micro-data, and monetising it,” one of the people quoted earlier said, adding that “the eventual goal is for it to go to external clients, and be leveraged with third-party players.”
This means that over time, Uber could package this labelled data—or the insights derived from it—into a commercial service for external clients such as enterprises, AV developers, logistics firms, or city agencies seeking ground-truth, frequently refreshed geospatial datasets. This could be channelled through Uber’s AI Solutions unit, which already offers data-labelling services to third parties.
Extracting more
Uber’s push follows its US arm’s acquisition of Belgium-based Segments.aia data-labelling startup, in early October this year as platforms with large physical touchpoints are shifting from passively collecting micro-data to actively curating and monetising it.
“For a mature platform like Uber, growth is now less about adding rides and more about extracting value from the same network,” said Farheen, analyst at the Centre for Critical and Emerging Technologies. “If drivers can use downtime to gather and label data, the company gets more out of every minute they’re on the clock.”
From service to information work
Uber’s move reflects a shift across the internet economy, where the line between service and information work is blurring.
Telecom operators Bharti Airtel monetizes consented subscriber data through Airtel ads, serving precision-targeted campaigns to 320 million users. MapmyIndia has turned its detailed geospatial datasets into revenue by licensing APIs (tools that allow two software to communicate) for navigation and analytics. While Delhivery’s OS1 platform converts logistics data from billions of deliveries into address-validation and routing services, Ola leverages daily terabytes of platform data for pricing, route optimisation, and personalisation.
Globally, Tesla relies on human annotators to review car footage; Amazon pioneered micro-tasking with Mechanical Turk; and JD.com tightly integrates AI into warehouses.
According to CUTS, India’s annotation economy is already worth ₹2,000 crore and projected to grow to more than ₹4,000 crore in tools alone by 2030 at about a 29% CAGR. India accounts for about 7.9% of the global market, with about 70,000 annotators (about 50,000 freelancers, 20,000 full-time). It remains heavily export-oriented, with around 60% of revenues coming from US clients.
Commercial calculus
“Uber clearly sees commercial upside—its AI Solutions unit already sells labelling services to third parties,” said Sohom Banerjee, senior research associate at CUTS International. “But the catch is labour economics. Turning downtime into gigs can top up driver income, but it also risks deepening precarious piece-work—millions of cloud workers in the Global South already label data for low pay with thin protections.”
Farheen flagged a tradeoff. “For drivers, it means their job keeps expanding while the rewards might not,” she said. Every new micro-task extends the platform’s reach without hiring more people, she said.
In FY24, Uber India reported ₹3,860 crore in revenue with a ₹89 crore loss, maintaining its lead over rivals. Ola’s ride-hailing arm posted ₹1,761 crore in revenue with a ₹10 crore loss, while Rapido logged ₹648 crore in revenue but a much steeper ₹371 crore loss.
Worth the effort?
Not all experts are convinced. “Google and Apple control foundational map ecosystems with satellites, proprietary sensors, and decades of behavioural data,” said Abhivardhan, president of the Indian Society of Artificial Intelligence and Law. “Uber’s driver annotations are incremental inputs, not architectural advantages. This is cost arbitrage on labour, not technical differentiation.”
Banerjee underscored the challenge, “Even after spending half a billion dollars building its own maps to reduce dependence, Uber still lags what’s effectively a decade-long head start by Google in scale and detail.”
In India, foreign mapping rules—capping resolution at one metre and pushing for on-shore storage via local partners like MapMyIndia—limit any proprietary edge. Uber’s integration with open-network policies such as ONDC further undercut exclusivity, as they are interoperable by design.
