India needs a human- and farmer-centred model of agricultural AI

Artificial intelligence (AI) is entering Indian agriculture at a time when the sector is already undergoing profound change. Weather uncertainty is increasing, farm sizes remain small, markets are becoming more integrated, and farmers are being asked to produce more with increasingly constrained land and water resources. AI promises to help address several of these challenges—from predicting weather and crop diseases to improving input use, market intelligence and supply chains. But there is a question that receives far less attention: if AI makes Indian agriculture more productive, who will capture the gains?

This is not a philosophical question. It is an economic one, and it could become one of the defining questions for Indian agriculture over the next two decades.

Much of the current discussion focuses on what AI can do for farmers. AI can process vast amounts of weather, satellite, soil, crop and market data and turn them into recommendations on sowing, irrigation, fertiliser application, pest management and harvesting. In livestock and fisheries, similar technologies could improve disease surveillance and productivity. Yet productivity is only the beginning of the economic story.

Suppose an AI-enabled advisory system increases crop yields. The farmer’s income will depend not only on the additional output but also on the cost of adopting the technology, the price at which the output can be sold and the farmer’s bargaining position. If thousands of farmers simultaneously increase production of the same crop, prices could fall. If the technology requires proprietary inputs, part of the gain could accrue to input suppliers. If a platform controls market information and access to buyers, another part could be captured further downstream. Thus, the economic chain is not simply AI leading to higher productivity and higher farmer incomes. Improvements in productivity affect production costs and output, which influence market prices, bargaining power across the value chain and ultimately the distribution of economic surplus. It is this distribution that will determine how much of the AI dividend actually reaches farmers.

This distinction needs to be at the centre of India’s AI-and-agriculture strategy.

AI divide in Indian agriculture

There is a potentially transformative opportunity because agriculture has traditionally suffered from significant information gaps. Farmers make production decisions with imperfect information about weather, pest incidence, prices, demand, input quality and market conditions. AI can reduce some of these gaps. But here lies a paradox: AI could democratise information—or concentrate it.

A farmer receiving a localised weather forecast through an affordable public platform gains better information for making decisions. But a large commercial platform possessing much richer information about crop conditions, farmer behaviour, local prices and future demand may acquire an informational advantage over the farmer. The same technology that reduces one information gap could create another.

This makes agricultural data particularly important. Indian agriculture is generating increasingly valuable data through land records, crop surveys, weather observations, satellite imagery, digital marketplaces, credit, insurance and government programmes. As AI systems become better at combining these datasets, questions of who controls, accesses and derives value from agricultural data will become increasingly important.  These are not merely questions of digital governance. They are questions about economic rents and bargaining power.

India’s experience with digital public infrastructure provides an important starting point. Open and interoperable digital systems can lower entry barriers and prevent essential infrastructure from being controlled by a handful of players. A similar principle could guide agricultural AI.

The objective should therefore be to create an agricultural AI ecosystem, rather than merely an AI market. Publicly governed datasets, digital platforms that can communicate with one another, affordable computing, common standards and open interfaces could allow startups, universities, farmer organisations and state governments to develop applications without recreating the underlying infrastructure.

Farmer Producer Organisations, cooperatives and other agricultural institutions can also play a greater role. They can help small farmers access new technologies, share information and negotiate better terms. Collective action can make AI more affordable and strengthen farmers’ bargaining power as agricultural value chains become increasingly technology-driven.

This matters because Indian agriculture is dominated by small and marginal farmers. The economics of AI developed for a large commercial farm cannot simply be transferred to a small holding. The technology must work within the realities of fragmented holdings, variable connectivity, limited liquidity and diverse farming systems.

There is therefore a danger of creating an AI divide within agriculture. Better-capitalised farmers and commercial enterprises may adopt sophisticated technologies early, while smallholders remain dependent on conventional extension systems. Over time, differences in access to information could translate into differences in productivity, market access and income. The answer is not to slow technological adoption, but to ensure that it is accompanied by complementary investments. AI cannot substitute for irrigation, storage, affordable credit or rural roads. An accurate recommendation has little value if a farmer does not have the means to act upon it. The economic returns to agricultural AI will therefore depend heavily on the infrastructure and institutions surrounding it.

Agricultural extension will also need to evolve. AI could generate highly localised information while human professionals help farmers interpret it, assess trade-offs and act upon it. The human interface will remain important because agricultural decisions depend not only on information but also on resources, risk preferences and local circumstances.

AI systems also need to be evaluated carefully. Indian agriculture is extraordinarily diverse across crops, regions, soils, irrigation regimes and farming practices. A model that performs well in one agro-climatic zone may not perform equally well elsewhere. AI-generated advice must therefore be tested against actual farm outcomes. This points towards an important principle: measure impact at the farm level, not merely adoption at the platform level.

The relevant measures should include changes in yields, input costs, water use, crop losses, price realisation, income volatility and, ultimately, farmer incomes. The number of farmers registered on an AI platform is not an impact indicator.

Key takeaways

Looking towards 2047, India should therefore consider alternative AI-enabled agricultural futures. In one, AI becomes an inclusive productivity technology: public infrastructure keeps access affordable, farmer organisations strengthen bargaining power, and productivity gains translate into higher and more stable farm incomes. In another, a few platforms control data, market access and agricultural intelligence, while farmers capture a smaller share of the value created. A third possibility is a two-speed agricultural economy in which commercially viable farmers adopt AI rapidly while millions of smallholders remain outside the technological frontier.

India’s policy challenge is therefore not simply to promote AI in agriculture, but to shape the economic architecture within which it operates.

First, agricultural AI should be treated as an element of public infrastructure. Government can create interoperable datasets, common standards, testing facilities and affordable access to computing. The objective is not for the state to develop every AI application, but to reduce the cost of innovation and prevent essential infrastructure from becoming excessively concentrated.

Second, farmers need greater agency over agricultural data. Data generated through farming activity should not become an invisible source of rents for technology platforms. Farmers and their organisations need meaningful rights over how their data are accessed and used, while appropriate safeguards should allow anonymised and aggregated data to support wider innovation.

Third, farmer organisations should become participants in the AI ecosystem. Farmer Producer Organisations, cooperatives and other agricultural institutions can help small farmers access new technologies, share information and negotiate better terms. Collective action can reduce the cost of adopting AI and strengthen farmers’ bargaining power.

Fourth, AI interventions must be evaluated against economic outcomes rather than technological adoption. The relevant question is not how many farmers use an AI application, but whether they experience higher net incomes, lower input costs, reduced crop losses, better price realisation and lower income volatility. Public policy should reward demonstrated outcomes rather than the proliferation of pilots.

This is ultimately a question of institutional design. Markets will provide powerful incentives for AI innovation, but markets alone cannot guarantee that productivity gains will be equitably distributed. Conversely, excessive state control could suppress the innovation that makes AI valuable.

India therefore needs a human-centred and farmer-centred model of agricultural AI: competitive enough to encourage innovation, open enough to prevent technological concentration, and institutionally strong enough to ensure that farmers participate in the value created from their data and adoption of new technologies.

The prize is considerable. AI could help Indian agriculture move from reactive farming to predictive farming; from generic extension to personalised advice; from fragmented markets to better-integrated value chains; and from managing risk after it occurs to anticipating it before it does. But the ultimate measure of success will not be the sophistication of the algorithms. It will be whether the farmer is better off.

India should therefore enter the AI era in agriculture with a clear economic bargain: technology may transform how agriculture is produced and marketed, but the productivity dividend must translate into better incomes, greater resilience and stronger bargaining power for those who produce our food.

That is the agricultural AI future India should seek to build by 2047.

Views expressed are the author’s own and don’t necessarily reflect those of Down To Earth