AI Monsoon Forecasts Reached 3.88 Crore Indian Farmers in 2025, the Kisan e-Mitra Chatbot Has Answered Over 93 Lakh Queries, and 10.31 Crore Farmers Now Hold a Farmer ID

The documented uses of AI in Indian farming include weather forecasts, chatbots, pest identification from photographs and yield estimates for crop insurance. One of them has already reached 3,88,45,214 farmers (38.8 million): the monsoon-onset forecasts that the Ministry of Agriculture and Farmers’ Welfare sent by SMS in 13 states in Kharif 2025, the rainy-season crop. The others are a chatbot that has answered more than 93 lakh (9.3 million) farmer queries, a pest-photo system used by more than 10,000 extension workers, a farmer registry with 10.31 crore (103.1 million) IDs that the AI services draw on, and two older pilots in Telangana and Andhra Pradesh. Each result below comes with the body that measured it and what that count leaves out.

A monsoon forecast on 3.88 crore phones

The Ministry’s release of 12 September 2025 says the forecasts went out through m-Kisan, its SMS service, up to four weeks before the rain. It rounded the reach to nearly 3.8 crore farmers. The Lok Sabha reply of 17 March 2026 gives the exact count and says the messages were in Hindi, Odia, Marathi, Bangla and Punjabi. The method was a blend of two AI forecast systems, Google’s NeuralGCM and the European weather centre’s AIFS, with 125 years of rainfall records from the India Meteorological Department (IMD). The Development Innovation Lab-India worked on the pilot with the Ministry.

The Ministry says the forecasts correctly flagged a 20-day pause in the monsoon’s northward advance that year, and that updates went out every week until continuous rain reached an area.

Telephone surveys run through Kisan Call Centres in Madhya Pradesh and Bihar found that 31 to 52 percent of the farmers asked had changed their planting decisions, mostly land preparation and sowing time, according to the Lok Sabha reply. The answers are self-reported, they come from two states, and the reply does not give the number of farmers surveyed.

A preprint of 9 March 2026 by Colin Aitken, William Boos, Michael Kremer and colleagues, who developed the system, tests the method on the years 2000 to 2024. Pooled across lead times up to four weeks, it cut the Brier score, an error measure for probability forecasts, by 5 to 10 percent against a plain historical-average forecast, and by about 20 percent in 2025. The authors list two caveats of their own: the AI models’ training years overlap some test years, and only 28 grid cells received forecasts in 2025.

The government said in March 2026 that it would move the capability into a national system built with the Indian Institute of Tropical Meteorology (IITM), IMD and ISRO, with IITM’s models added for 2026. On 12 May 2026 IMD launched its first AI-driven monsoon advance forecast, issued every Wednesday up to four weeks ahead for farmers in 16 states and more than 3,000 sub-districts. The scores of these and other AI weather models are in tuput’s piece on AI weather forecasting.

Kisan e-Mitra and 93 lakh answers

The Ministry launched Kisan e-Mitra in September 2023 to take PM-KISAN questions about payments, registration and e-KYC off the desks of block and district officials, Wadhwani AI writes. Wadhwani AI, a nonprofit, added its AI features in February 2024 with support from Google.org and in collaboration with Bhashini, EkStep, NIC and Samagra. The chatbot now works in 11 languages using Bhashini’s IndicTrans2 translation model, which tuput’s Bhashini article describes, and answers from preset templates filled with the farmer’s own record. Farmers can type, tap a common question or speak.

The Ministry’s backgrounder of 14 February 2026 says that by December 2025 the chatbot had answered more than 93 lakh queries and was handling more than 8,000 a day, covering PM-KISAN, the Kisan Credit Card and the crop insurance scheme PMFBY.

Wadhwani AI reports that between 6 March and 5 August 2024 the chatbot answered 2.69 million queries in 186 days, about 13,000 a day, which it calls a 668 percent increase over manual grievance handling. The builder counted its own tool, and a count of answered queries does not show how many farmers received the payment they had asked about. The Ministry’s daily figure was 20,000 in a Rajya Sabha reply of 28 March 2025 and 8,000 in December 2025, and the releases do not explain the change.

Bharat-VISTAAR, a helpline that talks

Phase 1 of Bharat-VISTAAR launched on 17 February 2026, with a Rs 150 crore allocation in the Union Budget 2026-27. A farmer can call 155261, use a voice chatbot, the Ministry’s web portal or a mobile app, and people without smartphones can reach it through an interactive voice response system, says the Ministry’s reply of 13 March 2026. Phase 1 runs in Hindi and English and brings in ten central schemes, among them PM-KISAN, PMFBY and the Soil Health Card, with crop practices from ICAR, the Indian Council of Agricultural Research. The Ministry’s pre-launch release lists IMD weather, AgMarkNet mandi prices, the National Pest Surveillance System and AgriStack data as inputs.

The releases give no usage figure for it yet, and they do not say which speech technology it runs on. Maharashtra’s own MahaVISTAAR-AI app, which Chief Minister Devendra Fadnavis inaugurated according to Elets eGov on 24 May 2025, is built on several public systems, including Mission Bhashini. That report gives no user numbers.

Pest photographs: NPSS and ICAR’s AI-DISC

The National Pest Surveillance System (NPSS), launched in 2024, lets a farmer or extension worker upload a photograph of a crop pest or disease and get a diagnosis and advice. The February 2026 backgrounder says that by December 2025 it was used by more than 10,000 extension workers and covered 66 crops and more than 432 pests. A March 2025 reply put it at 61 crops and over 400 pests, with about 1 lakh uploaded images.

ICAR’s Indian Agricultural Statistics Research Institute built AI-DISC, a phone app that identifies crop diseases and pests from a photograph. The institute’s own achievements document says it detects 55 diseases in 22 crops and 66 insect pests in 8 crops, has more than 10,000 downloads on Google Play, and shares its models as APIs, the software links apps use to talk to each other, with NPSS and a second government project called POCRA.

Neither source measures accuracy or crop saved. The NPSS figures are counts of users, crops and uploads, and the ICAR document lists coverage and downloads. It names the cost of computing and the need for skilled staff as constraints, and asks for a multi-institution project to cover more crops.

AgriStack, the registry under the tools

AgriStack is a database, not an AI tool, but the AI services are being connected to it. Its Farmer Registry gives each landholding farmer a Farmer ID linked to land records. The Ministry’s release of 7 August 2026 says more than 10.31 crore Farmer IDs had been created by 3 August 2026, and that the Digital Crop Survey covered more than 31.3 crore plots in 648 districts in Rabi 2025-26. The February 2026 backgrounder gives the target as 11 crore IDs by 2026-27. States create and hold the data, and the Ministry compiles the counts.

The Ministry’s release of 24 March 2026 reports two uses. Maharashtra transferred over Rs 14,000 crore to 89 lakh farmers for Kharif 2025 crop losses within five days, and Chhattisgarh used Farmer IDs and the crop survey for paddy purchase at the minimum support price from more than 32 lakh farmers in one season. These are the Ministry’s reports of state results, and the release cites no outside evaluation.

Under PMFBY, a tool called YES-TECH uses remote sensing and AI analytics to estimate yields for paddy and wheat from Kharif 2023 and for soybean from Kharif 2024, with at least 30 percent weight on technology-based assessments. The backgrounder says that as of January 2025 nine states had adopted it and Madhya Pradesh was moving fully to technology-based estimates.

Saagu Baagu, chilli farmers in Khammam

The World Economic Forum’s AI4AI initiative, led by its Centre for the Fourth Industrial Revolution India, ran Saagu Baagu in Khammam district, Telangana, with the state government, Digital Green as implementer and the Gates Foundation as supporter. The Forum’s playbook says four AI-enabled applications went to around 7,000 chilli-growing smallholders.

The Forum’s article of 26 February 2024 reports first-phase results for participating farmers: yield per acre up 21 percent, pesticide use down 9 percent, fertiliser use down 5 percent, unit prices up 8 percent, and income up by more than Rs 66,000 (about $800) per acre per crop cycle. The February 2025 playbook, written with BCG X, puts the price gain at 11 percent and the cut in fertiliser and pesticide use together at 9 percent, and says the farmers reported the 21 percent yield gain after one season. The 2024 article does not say who measured the results or how, the playbook attributes them to the farmers’ own reports, and neither describes a comparison group.

In October 2023 Telangana widened the project to five crops in ten districts, with a target of 500,000 farmers.

ICRISAT and Microsoft’s sowing advice in Andhra Pradesh

In the 2016 rainy season ICRISAT, Microsoft and the Andhra Pradesh government sent ten SMS advisories in Telugu to 175 groundnut farmers in Devanakonda, Kurnool district. The sowing window, which opened around 24 June, came from 30 years of climate data (1986 to 2015) and a soil-moisture index.

ICRISAT’s note of 13 January 2017 says registered farmers’ yields were 30 percent higher than those of some non-registered farmers who had sowed in the first week of June, based on crop-cutting experiments at selected plots. That June-sown crop met a dry spell of about 8 mm of rain over 31 days from 10 August 2016, so the gap reflects one season’s weather. The note does not give the size of the comparison group or the raw yields.

Microsoft’s release of 7 November 2017 reports an expansion in 2017 to more than 3,000 farmers in Andhra Pradesh and Karnataka, with gains of 10 to 30 percent across crops. The article gives no method for those figures.

Drones, and what comes next

Spraying drones are a separate scheme, covered in tuput’s Namo Drone Didi article. The Ministry’s reply of 13 March 2026 says Bharat-VISTAAR will add four languages within three months and five more within six months.