An AI Revolution in Agriculture: Is it around the corner?
Artificial Intelligence or AI is clearly the flavor of the season. AI has probably dominated at least half of all articles that I have read over the new year, on what to expect in 2025. In the field of agriculture too, there have been some very innovative AI enabled products; such as an ‘Image Recognition’ app that helps farmers diagnose pests and diseases by just taking a photo, another application uses a similar algorithm to visually assay grain quality before they are marketed or an applications being developed by a foundation that helps cotton farmers predict a boll worm attack. The Government of India has also rightly entered this fray, with initiatives on to develop an Agri Stack and use AI to provide localized and custom level advisory to farmers.
At the same time, experience shows that Indian farmers in general have been reluctant to adopt such digital solutions. And no, it is not because of lack of training or awareness. A digital revolution that has swept most parts of our lives today, is yet to take off in Agriculture. Does the same fate await AI products?
There is no easy answer to why farmers won’t readily adopt digital (and AI) based products. But some observed patterns may be instructive. To start with, I have seen many tech entrepreneurs (and presumably their investors) start with two assumptions — (i) There is a huge information gap in the rural areas and if given a digital tool to bridge this, farmers will lap it up and (ii) What makes perfect economic sense to me, will do so for the farmer as well. These assumptions put them on a wrong foot right from the beginning. I have seen such enterprises struggle to onboard farmers and/or see sustained usage of these services. Often there is a huge drop in usage within the very first month itself. Below, I argue that three factors that need to be considered when designing any digital information product for farmers.
1) How much is Agriculture a priority for small farmers in India? And if it is not, how much mind space are they willing to dedicate to study analytical data thrown at them and accordingly undertake intensive interventions on their farms. This question may seem counter intuitive, but anyone who is working with small and marginal farmers will tell you that, farming accounts for a small fraction of the total household incomes. And these small and marginal farmers account for more than 86% of total holdings as per the agriculture census. I recall a survey of farmers from Haryana and Rajasthan that I had done in 2018, for a client I was advising. The study indicated that incomes from farming constituted only 20% of total household incomes of small and marginal farmers. Some simple back of the envelope calculations further corroborates this. The weighted average holding size of small and marginal farmers (as per the census) comes to 0.5 Ha. Assuming say such a farmer grows an oilseed like Soyabean in Kharif and chickpeas in Rabi. He is most likely to earn a net income of around Rs 50,000 per annum from two crops combined, which is small even for a rural household. Comparatively, the same farmer, would earn almost double, if he and his spouse offer labour on a large farmer’s land for 8 months in a year, and fill up the rest through MGNREGA.
The point I am trying to make is that, if farming is not a top income priority for more than 85% of farming households (who probably consider output from farming either as a subsistence or at best some additional incomes), they will be unwilling to invest time and effort to study analytical data of their farms and take intensive interventions. Among the information overload that all of us undergo (including farmers), this data is most likely to be ignored. And it is even more unlikely to be able to monetize this service with payments expected from farmers.
2) How does a farmer’s make economic decisions? I have realized that many people fail to understand this, despite spending decades working in agriculture. Let me cite an experience I had in Karimnagar (Telangana) during my early days in the agri sector. It taught me a lot about how a farmer makes his economic decisions. I observed a farmer whose cotton farms were full of weeds, but he couldn’t care two hoots about it. It was a drought prone year, and his cotton crop was already in distress. While my logic dictated that he should do something about the weeds, for the farmer the cost, time and effort to de-weed the farm was more than the output he was realistically expecting from the field. Professionals working in the agri extension field would identify a common response from farmers , whenever they are taken to a demonstration plot (by say a university or a KVK), that “this is too expensive (and risky) for me to replicate”. Our perceived cost benefit assessment of a farm often fails to factor in factors like probability of success of the entire crop, given external risk factors (primarily climatic) and its economic impact on the initial investment. Farmers are often reluctant to make an upfront heavy money or time investment in their farms, unless they are reasonably assured of success. Many farm AI services hinge on use of expensive IoT devices, or use of Drones and/or satellite data for input to their AI models. A farmer is unlikely to make such investments upfront. And unless he is reasonably assured of high returns, he is unlikely to invest a lot of time in undertaking a wide variety of new and intensive interventions in his farm.
3) Comfort Factor. A common refrain in the farming sector is that children of farmers do not want to do farming. Most youth from farming families aspire for non-agri careers. Not surprisingly, most people who are manning farms today are in their middle ages or older. And many of them lack a high degree of comfort and trust in new technologies. The traditional way of accessing new knowledge has been either through peers, the agri-input shop or the occasional KVK scientist of NGO workers who comes your way. The challenge with information driven through an app are two-fold — (i) Can he trust this information and (ii) whom does he catch hold of, if the information was wrong. This breaking of barriers to seeking information digitally is going to take time.
The above factors remind of one of C K Pralhad’s key arguments in his seminal book -” Fortune at the bottom of the Pyramid”. I am paraphrasing; he said that serving the bottom of the pyramid does not mean rehashing existing products to develop cheaper versions. It rather means to develop products especially for this clientele, keeping in their mind their unique needs. There couldn’t be a more valuable advice for tech entrepreneurs. That said, below are my two pennies worth on what could be useful additions from AI to the Agri sector in India.
1) Give Solutions not Data. To take an analogy from the stock markets, it does not help me if you just tell me the PE ratio of a stock, unless you give me information on how to interpret it and make a clear recommendation. Similarly, what farmers possible need are not data but clear recommendations. Add a disclaimer, if you must, but provide a recommendation.
2) Practical Suggestions. I have always been a great believer in the 80:20 principle. What are the 20% things that bring 80% impact? AI tools providing ultra-local and custom advisory should make it easy for the farmers to follow their advice. Suggesting 2–3 high impact interventions (that are also low hanging fruits) can not only help nudge farmers towards changing their behaviors but also help build trust over time. Someone very rightly said once, building trust is a process and not a product. That should be an inherent part in any product journey.
3) Think of the intermediaries. AI and Technology products are often so obsessed with targeting the farmers that they forget the many intermediaries — such as NGOs, KVKs, Agri department staff etc. — who act as an effective last mile to farmers. Can AI tools be developed to make their jobs more effective. For instance, instead of giving personalized analysis to a farmer for his farm, can a more localized area specific analysis and recommendation be given to the extension worker of an NGO, who can then appropriately translate the same to the farmer? I believe that will set a more sustainable template for AI adoption.
I do believe that AI has a lot to offer to the agriculture sector in India and its myriad set of challenges. But we need to think about what solutions are we developing and what is the most effective way of taking them to the last mile.
