From cow collar to crop advisory: How AI is slowly entering India’s farms

In Maharashtra, a collaboration between Agriculture Development Trust (ADT) Baramati, Microsoft, and Oxford University has used AI-enabled crop management to improve sugarcane yields by 25 to 30 per cent. Work is now underway to extend similar interventions to vegetables such as tomato and brinjal, with algorithms being developed to bring tailored advisories to individual farmers.

“What we are doing in sugarcane – soil health management, soil moisture, soil nutrients, crop health, crop nutrition, and crop protection – similar things will be done for vegetables,” said Dr Bharat Kakade, President and Managing Trustee, BAIF Development Research Foundation, Pune.

The AI-enabled neck collar fitted on a cow at the Centre of Excellence in Debari, Udaipur, monitors animal health, nutrition, and heat cycles in real time. (Express Photo)

Another area generating interest is weed management. Currently, farmers either pull weeds manually, which is labour-intensive and expensive, or spray herbicides that carry risks to human and animal health. AI-powered robotics offer a third path: machines that move through fields and remove weeds without chemicals. The technology is in an early stage, but Kakade sees it as one of the more promising frontiers. “AI-enabled robots can be sent into farms to do weeding, as an alternative to herbicides, which have hazardous effects on human health,” he noted.

Government schemes

The Digital Agriculture Mission, launched in 2024 with a total outlay of Rs 2,817 crore, aims to advance farmer-centric digital solutions by leveraging verified datasets on farmers, landholdings, and crops, alongside AI and remote sensing technologies.

At its core is AgriStack, a digital identity system linking each farmer to their land records, livestock, and government benefits. Over 7.63 crore Farmer IDs have been generated as of November last year, including 1.93 crore for women farmers, against a target of 11 crore by 2026-27.

Looking ahead, the Union Budget 2026-27 has proposed Bharat-VISTAAR (Virtually Integrated System to Access Agricultural Resources), a multilingual AI tool that aims to integrate AgriStack portals and Indian Council of Agricultural Research (ICAR) data to provide customised, voice-first, AI-driven advisory support to farmers to enhance productivity.

Early tools

Two national tools stand out for their reach. Kisan e-Mitra, an AI-powered chatbot operating in 11 regional languages, handles over 8,000 farmer queries daily, covering schemes such as PM Kisan, Kisan Credit Card, and crop insurance. The National Pest Surveillance System (NPSS), launched in 2024, allows farmers to upload images of affected crops for rapid AI-powered diagnosis.

For sowing decisions, an AI-based pilot for local monsoon onset forecasting during Kharif 2025 reached 3.88 crore farmers across 13 states via SMS, with 31 to 52 per cent of surveyed farmers adjusting their sowing decisions based on these forecasts.

Protecting what farmers grow

Crop losses have long been financially devastating for small farmers. AI is now making insurance faster and more transparent. YES-TECH (Yield Estimation System based on Technology) uses remote sensing and AI-driven analytics for yield estimation and has been adopted by nine states, with Madhya Pradesh fully transitioning to technology-based yield assessment.

CROPIC (Collection of Real-Time Observations & Photo of Crops) allows farmers to upload geotagged, time-stamped crop images to support transparent, real-time damage assessment under the Pradhan Mantri Fasal Bima Yojana.

Challenges

Despite the momentum, the technology is far from ready to operate independently. Kakade noted that with datasets still being built in many areas, AI recommendations cannot yet be fully trusted without human verification alongside them. “Unless you have sufficient data collected, gathered, analysed, and until accuracy reaches a certain level, you can’t rely on solutions. As data increases, accuracy and reliability improve,” he said.

In a sector where a wrong advisory can mean a lost harvest, AI outputs at this stage must be read alongside the judgment of trained field workers and extension officers.

The government’s own AI Playbook for Agriculture, developed with the World Economic Forum, acknowledges that fragmented data ecosystems, limited digital infrastructure, affordability barriers, and last-mile delivery challenges remain critical constraints to scaling AI from pilots to widespread adoption.

“Skilled manpower at the intersection of agriculture and AI is also limited, and funding support in the early stages remains necessary before solutions reach viability,” added Kakade.