Land registries and infrastructure: AI where the state is fragmented
In India, at least half a dozen start-ups use artificial intelligence to map road damage and report it to the authorities, according to the BBC. It is one strand of a broader phenomenon: young companies applying AI to typically Indian problems, in sectors ranging from legal services to healthcare, and from infrastructure to education and agriculture. The common thread, according to the BBC, is working with highly fragmented and multilingual data in markets that are highly price-sensitive.
The most frequently cited case is Landeed, a platform founded by Sanjay Mandava after a personal dispute over family land in Chennai. In India, the founder tells the BBC, a system of presumptive title applies: it is the buyer who must prove ownership, not the state that certifies it as in the United States or the United Kingdom. Hence the need to aggregate different documents — sale deeds, building certificates, tax receipts, cadastral maps — and verify inconsistencies. The platform, active in 24 states, claims an archive of 1.8 billion documents and a reading and checking process that would take lawyers and intermediaries days.
On the road front, Bengaluru YouTube educator Gaurav Sen has developed an app that detects potholes using dash-cams and computer-vision models, identifies the relevant offices and sends them reports. Sen stresses to the BBC that the app can be used by municipal companies on public buses and taxis to obtain up-to-date data in real time. It is an example of how public demand, rather than frontier research, is steering adoption.
Healthcare and funding: growth and its limits
In Mumbai, Fractal Analytics has installed the "Vaidya" tool in hospitals run by the local civic authority. According to the BBC, the app offers citizens simplified summaries of laboratory reports in multiple languages, points users to nearby healthcare facilities according to their needs and provides doctors with pre-consultation support including clinical history and treatments. For administrators, a public-health dashboard based on aggregated and anonymous data is planned, useful for identifying emerging infections or epidemiological trends.
The sector is attracting capital: according to the platform Inc42 cited by the BBC, fundraising by Indian AI start-ups grew 265% year on year, surpassing $1 billion in the first nine months of the year, in an overall weak fundraising environment. The BBC, however, frames the phenomenon as primarily application-driven: unlike the United States and China, which dominate frontier language models, India is moving in the areas of data infrastructure and consumer applications.
The bottleneck: computing, chips and bureaucracy
The difficulties reported by the founders are concrete. Srikanth Velamakanni, co-founder of Fractal, observes that the computing power required for a model is the same everywhere, but the spending power of American clients is far higher than that of Indian clients: building in India, he says, sometimes costs more than clients are willing to pay. Mandava tells the BBC he is "short of GPUs" and has enormous difficulty buying them. The government has promised 38,000 GPUs to start-ups through its national AI mission, but demand exceeds supply.
Regulation and bureaucracy also weigh on the sector, along with the difficulty of turning successful pilots into recurring public contracts and dependence on costly foreign technology. The BBC reports experts' fear that, without resolving these issues, Indian companies will produce superficial solutions that are quickly overtaken. Sen himself admits that his product, if presented to a major player such as Google, would make people smile, and that to be sustainable it will have to become more sophisticated. It remains to be seen, over time, whether funding is actually converted into revenue and what the real impact of public promises on computing will be.
