The policy push is already visible. The Union Budget 2026-27 has positioned AI as a key component of agricultural transformation in the country, with AgriStack functioning as core digital public infrastructure.
For the uninitiated, AgriStack is a centrally designed, state-executed system where the Centre sets standards, architecture, APIs and privacy rules, and the states build farmer, land, and crop registries using district-level data. Data stays with states and is shared via consent-based, federated APIs, with platforms like AgriKosh linking research and transaction data. Services such as subsidies, credit, insurance, advisories and markets run on top, supported by AI that works even on low-bandwidth or offline devices.
The government even highlighted that AI is steering Indian agriculture towards a data-driven, farmer-centric, and sustainable model during the recently concluded IndiaAI Impact Summit 2026 in New Delhi.
However, its success, experts say, hinges on granular, reliable data, which is still limited by fragmented landholdings, low digital literacy, and weak data systems. Such limitations may pose challenges to realising the full potential of AI in agriculture. Without addressing these issues, they say AgriStack may not be able to achieve its objectives.
Area under crops in India (thousand hectares)
Source: Land Use Statistics for 2022-23, Agriculture Census Unit, DA&FW“What is missing is a strong middle layer for governance and decision-making, which creates gaps in coordination and accountability. Agriculture still lacks systems that ensure end-to-end accountability and traceability—capabilities that have been successfully established in platforms like the Unified Payments Interface (UPI),” says Gopal Krishna Patra, Director, CSIR-Fourth Paradigm Institute (CSIR-4PI), Bengaluru.
These challenges become even clearer when viewed against the country’s extensive and diverse agricultural terrain. India holds the second-largest agricultural land in the world, with 179.98 million hectares (as per the annual report 2024-25 by the Agriculture Ministry) used for farming. This vast expanse is mainly spread across diverse land types, including irrigated land, rainfed/dryland, and plantations, as well as various soil types, such as alluvial soil in the Indo-Gangetic plains, black soil in the Deccan region, and red and laterite soils in the southern and eastern regions.
More than 86% of Indian farmers are small and marginal, with average landholdings just over a hectare in size, and they generally possess low levels of digital literacy. Experts say that all these factors, combined with a weak data infrastructure, have resulted in inconsistent datasets, leaving AI models ill-equipped to address India’s highly diverse, data-poor smallholder agriculture.
“Unlike the US or EU, where farms are sensor-rich and data systems are standardised, our smallholder systems are heterogeneous and data-poor,” says Patra. Due to these challenges, many AI tools remain stuck at pilots without multi-season, multi-location validation, he says. “Without field trials and farmer feedback, models risk being unreliable.”
So, agri-AI models must be region-specific and continuously recalibrated to reflect India’s diverse agro-climatic conditions, argues Patra. “For broader adoption, AI systems must be credible, authenticated, and demonstrably reliable.”
‘Customise models to local conditions’
Aditya Sesh, a member of the expert committee at the Ministry of Agriculture & Farmers Welfare, emphasises that regional relevance is vital for data capture, given that the majority of our farmers are smallholders and the agro-climatic diversity is vast. Unlike the advanced economies, where precision agriculture draws multi-billion-dollar private investment, “India’s agri-AI ecosystem remains largely public-led and still evolving.”
Details of important parameters of Land Use Statistics for the year 2022-23
Source: Land Use Statistics for 2022-23, Agriculture Census Unit, DA&FWNotably, India has built a large digital backbone for agriculture, with over 76.3 million Farmer IDs and 235 million crop plots mapped under the Digital Agriculture Mission. The Union Budget 2026-27 has allocated Rs 150 crore for Bharat-VISTAAR, a multilingual AI platform linking AgriStack and ICAR data to deliver personalised, real-time crop and risk advisories, further strengthening the digital agri infrastructure.
Agricultural AI must operate through a layered governance structure, believes Sesh, who is also the Founder and Managing Director of Basiz Fund Services. “The central stack, such as Bharat-VISTAAR, should define standards, interoperability, and secure architecture. To facilitate ingestion and reconciliation across ministries, states, and local entities, a middle layer should serve as a hub for structured data exchange. AI can eliminate inconsistencies, but governance needs to enforce outcome tracking and validation. Implementation must sit at the regional level, reflecting district-level conditions,” he says.
Accountability, Sesh says, must rest with a designated authority under the Ministry of Agriculture, measured by yield gains, cost savings, and income growth. Without clear metrics and ownership, even the best technology will fail to deliver farm-level impact, he notes.
Several farmers in the country also believe that AI in agriculture cannot be overly centralised because climatic and soil conditions vary widely across regions. Anil Ghanwat, President of Shetkari Sangathan, the largest farmer association in Maharashtra, says, “To truly help farmers, AI must provide guidance tailored to each farmer’s local ecosystem. AI can help analyse soil health and rainfall patterns, but for these insights to be useful, strong on-ground agronomic expertise is essential. Agronomists are needed to study regional conditions and translate AI insights into practical advice. In reality, the best solution for farmers is a combination of AI and skilled agronomists.”
“At the same time, awareness of AI remains low among small farmers. Increasing awareness and building trust will be critical for wider adoption. Nothing can work in isolation,” he adds.
Meanwhile, the Indian Council of Agricultural Research (ICAR) is digitising research and building interoperable datasets. However, experts contend that scaling from pilots to nationwide deployment remains uneven, especially in digitally weaker regions.
ICAR officials even acknowledge these challenges but say they are working to ensure that AI technologies are developed with small and marginal farmers in mind, while also accounting for local conditions. At the India AI Impact Summit 2026 in Delhi, Anil Rai, Assistant Director General (ICT), ICAR, indicated that ICAR is developing technologies specifically designed for small and marginal farmers, prioritising cost efficiency and data accuracy central to building AI models for agriculture. “With 113 institutions working across diverse agricultural domains, the government is undertaking extensive work in agricultural AI, particularly through ICAR. But innovation will be crucial in shaping the future of India’s farming sector,” says Rai while speaking to Economic Times Digital at the summit.
Missing middle layer
Not just innovation, the other key factor in shaping India’s agriculture is the middle layer. As AgriStack and state farm land registries expand, Patra asserts the need for a strong middle governance layer to connect national digital systems with local realities.
With the farmers spread across diverse agro-climatic zones, most models rely on datasets that are too aggregated, outdated, or disconnected from field realities, says Tauseef Khan, Co-founder of Unnati. “When a farmer receives a pest advisory calibrated for national-level averages, it’s not just unhelpful; it erodes trust in the entire digital promise,” he notes.
The middle governance layer must ensure interoperable data, contextual models, multi-season validation, and integration with KVKs, FPOs, cooperatives, and field staff, says Patra. Farm-level accountability should sit with a joint national-state body under the Ministry of Agriculture, tracking KPIs, such as yield, input efficiency, pest control, and income, he notes, saying a National Agricultural AI Mission Directorate, working with ICAR and state universities, should oversee audits and outcomes, turning AgriStack into a true decision-support backbone.
iStockThe middle governance layer must ensure interoperable data, contextual models, multi-season validation, and integration with KVKs, FPOs, cooperatives, and field staff, say experts.
“Ultimately, the middle governance layer must convert digital infrastructure into agronomic impact, measurable at the farm, season after season,” says Patra.
‘UPI model worth replicating’
India’s experience with the Unified Payments Interface (UPI) has shown that achieving trust at scale requires more than just technology; it needs robust institutional design, an interoperable architecture, and clear accountability.
Patra emphasised three essential points for the success of AI in agriculture. “Every advisory should carry a unique ID, model version, data source reference, timestamp, and confidence level. Data provenance and model lineage must be logged through immutable audit trails. Farmers and institutions should be able to answer a simple but powerful question: What data and which model generated this advice? Third, at the policy level, we must define liability, certification norms, and farmer data rights. AI advisories influence crop decisions, input costs, and incomes; therefore, accountability cannot be ambiguous. And mandatory validation across agro-climatic zones, periodic recalibration, and transparent consent frameworks are essential,” adds Patra.
Sesh says clear procedures for data exchanges, audit trails, and reconciliation checks must be institutionalised. “Over time, common reporting taxonomies, similar to financial standards, should be developed for agriculture. The ultimate interface for farmers will not be complex dashboards but trusted access points like Kisan BPOs, smartphones, and trained “Kisan buddies” at the grassroots. In the background, VISTAAR would make sure that the answers provided via these channels are contextually appropriate and supported by empirical data,” adds Sesh.
Agri-AI needs UPI-like traceability from advisory to outcome, say experts. “Just as UPI allowed any bank or fintech to build on its rails, India needs an ‘AgriAI Open Stack’ that allows private platforms to plug into government data (weather, soil health, and scheme eligibility) while maintaining quality standards,” says Khan of Gramophone.
“The Rs 60,000 crore Special Central Assistance package recently announced for AgriStack infrastructure is a transformative commitment, among the largest globally for agricultural digital public infrastructure. Building on this momentum, the policy framework can further strengthen the ecosystem by incorporating independent audits for model effectiveness, clear farmer data rights legislation, and outcome-linked performance standards for AI advisory platforms. This will ensure that India’s agricultural AI ecosystem earns and sustains the same level of public trust that UPI enjoys today,” he says.
Effectiveness of India’s AI models
Experts say India’s agro-climatic diversity makes agri-AI essential yet complex, requiring continuously recalibrated, not static, models. This demands integrated real-time data, farmer and extension feedback loops, region-specific modular design, and multi-season field validation with independent audits to ensure transparent, trusted advisories.
“As for recently launched platforms like AI VISTAAR, their long-term effectiveness will depend less on technological sophistication and more on governance. If they enable continuous learning, district-level adaptation, and measurable farm-level impact, they can be transformative. If they remain centralised advisory dashboards without validation and accountability, their impact will be limited,” says Patra.
Ashima Bajaj Seth, Chief Digital & Information Officer, Godrej Agrovet, says early evidence shows AgriStack-enabled AI advisories in India are directionally comparable to productivity and efficiency gains in developed economies.
“Global evidence indicates that precision agriculture and AI-driven solutions typically deliver 10-20% improvements in efficiency, yields or input optimisation, as highlighted in FAO’s case studies on digital and automation solutions in precision agriculture. The World Bank similarly notes that data‑driven digital agriculture consistently improves crop yields, reduces waste, and lowers input costs when farmers receive timely, localised recommendations,” adds Seth.
India’s early AI advisory pilots show similar directional gains, but from a lower base, making even small improvements highly income-positive for smallholders, say experts. However, matching the scale and durability seen in the US or Israel, they believe, will require much higher R&D investment.
“India currently spends 0.3-0.4% of its agricultural GDP on R&D, compared to about 0.7% in the US, based on comparative analyses from institutions such as IFPRI. Hence, to sustain the impact, India will need continued investment in AI model development, hyper-local datasets, climate analytics, and last-mile scalable digital extension frameworks, especially given our country’s diverse agro-climatic zones and smallholder-driven agriculture structure,” notes Bajaj.
iStockAgri-AI needs UPI-like traceability from advisory to outcome, say experts.
The global agri-AI market is projected to hit $5.9 billion in 2025, growing at nearly 26% CAGR. The IndiaAI Mission (Rs 10,372 crore), Digital Agriculture Mission (Rs 2,817 crore), and the recently announced Rs 60,000 crore Special Central Assistance for AgriStack signal intent and rank among the world’s largest commitments. Yet, given that agriculture contributes 18% to GDP and 42% to employment, there is still considerable potential for growth. For the context, more than 70% of large US farms currently use AI-based monitoring, while EU subsidies and regulation support widespread precision farming.
Since 2014, India’s agritech start-ups have raised over $2.4 billion, but most funding has flowed into marketplaces, not deep-tech AI. Experts say India needs a 5-10 times jump in investment to build India-specific AI models trained on local soils, crops, and weather rather than retrofitting Western models.
‘Need AI advisories built on ground realities’
Khan flags three gaps: farmer-level transaction data remain largely undigitised; 40-50% of agri-inputs are local or unbranded and outside formal systems; and counterfeits account for 15-20% of pesticides and nearly 10% of seeds, leaving AI advisories built on incomplete realities. With over 86% of farmers being smallholders averaging 1.08 ha and many using feature phones, reach is as critical as accuracy. India’s extension ratio is about 1:1,000 farmers versus 1:200-400 in the US and EU, making last-mile delivery decisive.
Patra says, “Effective agricultural AI requires far more than sophisticated algorithms; it depends on strong governance frameworks, common data standards, and robust digital infrastructure. As institutions like ICAR continue digitising research outputs and promoting interoperable datasets, the next wave of investment must focus on building systems that enable scale, particularly in digitally underserved regions.”
“First, India needs high-quality, continuously updated, district-level data ecosystems. Soil health records, crop-cutting data, weather streams, pest surveillance, and market information must be standardised, machine-readable, and interoperable across states. Equally important is data stewardship, with clear metadata standards, role-based access controls, and audit mechanisms to ensure trust and accountability,” he says.
Rural connectivity must be treated as critical agricultural infrastructure, say experts. AI tools cannot scale where broadband is unreliable, making investments in rural 4G/5G coverage, low-bandwidth and offline applications, and voice-based advisory systems in local languages essential for inclusive adoption.
“The computing infrastructure should not remain fully centralised. While national cloud platforms are necessary, decentralised or edge computing capabilities at the state or district levels can support faster model recalibration and context-specific analytics. Federated learning approaches can strengthen local adaptation while maintaining data privacy. Finally, sustained funding for ground-truth data collection and multi-season validation networks is indispensable. Without rigorous field validation and measurable outcome tracking, AI systems will remain pilots rather than trusted operational tools,” adds Patra.
“At the ecosystem level, strengthening last mile access to agricultural technologies is essential. India already has strong potential in terms of electrical and digital infrastructure, but the supply chain and service networks that support farm automation solutions must continue to expand. Companies with extensive distribution and service networks across rural India can play an important role in enabling faster and more reliable adoption of such technologies,” says Naresh Kumar, Chief Operating Officer, Lauritz Knudsen Electrical & Automation.
With supportive policy, farmer awareness, and enabling tech ecosystems, AI solutions can cut input costs, optimise water and energy use, and boost sustainable productivity, Kumar notes. “Such interventions have the potential to increase crop yields by as much as 30 to 40 percent, while also supporting resource efficiency and strengthening India’s long term food security,” adds Kumar.
Experts say ICAR’s scientific backbone must be linked with real-world transaction data via platforms like AgriKosh. Even as BharatNet expands connectivity and the IndiaAI Mission builds compute capacity, AI must work offline on low-bandwidth devices.
If investments in data, connectivity, computing, and validation align, stakeholders say agri-AI can shift from pilots to a scalable, farmer-centric system that lifts productivity, resilience and incomes.
At the AI4Agri 2026 Summit in Mumbai, Union Minister of Science and Technology and Earth Sciences, Jitendra Singh, noted that the country’s 140 million farm holdings, mainly small and marginal, could generate around Rs 70,000 crore annually if AI advisories enable each farmer to save just Rs 5,000 through optimised inputs, pest forecasting, and market linkages.
No doubt, the potential is enormous!