By Milind Borgikar, Co-Founder & CTO, Ayekart
Agriculture still employs about 43% of India’s workforce, according to an SBI Research analysis of Periodic Labour Force Survey data. Meanwhile, India has 958 million active internet users, 57% of them rural, according to the IAMAI–Kantar Internet in India 2025 report. It also finds that 44% of users already use AI features such as voice search, and nearly one in five goes online through someone else’s phone.
The story is clear. Rural India is online, comfortable talking to machines and often shares devices. Yet around 579 million Indians are still not active internet users, and the country speaks 22 scheduled languages and hundreds of dialects. The people who grow and trade India’s food are more reachable than ever, but most digital products still do not speak their language.
The next agritech revolution will not be won by the smartest algorithm, but by the most accessible one. It must be voice-first and inclusive by design.
The interface gap at the last mile
India has built remarkable digital rails. UPI, Aadhaar, account aggregators and affordable smartphones have changed what is possible in the village. Yet agritech adoption at the last mile remains stubbornly shallow.
The problem is rarely the technology underneath. It is the interface on top. Too many products are designed in metro offices for users who read English fluently, navigate dashboards with ease and trust forms. That is a small slice of the people who grow and trade India’s food.
According to the Agriculture Census 2015-16, about 86% of Indian farmers are small and marginal, cultivating less than two hectares. Many more are literate in Marathi, Odia or Telugu but are rarely addressed by software in those languages. When we ask them to adapt to our interfaces, we quietly exclude the very people agritech claims to serve.
Why voice, and why now
Rural India runs on conversation. Prices are discovered over phone calls, credit is extended on the strength of a familiar voice, and crop advice travels from neighbour to neighbour. Voice is not a feature to add later. It is the village’s native interface.
For years, speech recognition stumbled on accents, code-mixing and the noise of a tractor or a busy mandi. That is changing fast. Helped by open research efforts such as IIT Madras’s AI4Bharat, speech models now handle Indian languages far better than they did five years ago, including speakers who mix Hindi, English and a local dialect in one sentence.
Voice also fits practical realities. It works when hands are full of produce, is lighter on bandwidth than image-heavy apps, and through IVR can reach the basic feature phones still found in millions of rural homes.
The farmer is only one voice
The agrifood chain is long. An FPO manager logs procurement with a weighing scale on one side and a queue of farmers on the other. A small trader in a tier-3 town checks a payment while serving customers. A processor needs to know which producer groups can supply a given grade next week. Each would rather say what they need than navigate menus. AI’s job is to turn that spoken request into a structured transaction, so users never have to learn the system’s vocabulary.
Inclusion is an engineering decision
Voice alone does not guarantee inclusion. An assistant that only understands a standard urban accent simply moves the exclusion to a different layer.
Women are the clearest test. They do a large share of India’s farm work, yet are less likely to own the household smartphone or hold the land title. A system that assumes one registered user per farm will never hear them. Designing for shared devices, for women’s voices in training data, and for the questions that matter to them, such as livestock, post-harvest losses and household credit, is not charity. It is accuracy.
The same applies to tribal and older farmers, and to anyone whose dialect rarely appears in datasets.
India has the building blocks
India is treating language technology as public infrastructure. Bhashini, built under MeitY’s National Language Translation Mission, is creating open speech and translation capabilities across Indian languages. Combined with UPI, account aggregators and the emerging AgriStack, the pieces exist for a farmer to ask a question aloud and complete a real transaction. Private builders must now turn these rails into trustworthy experiences built on open standards.
Five principles for building for Bharat
- Start with the spoken question, not the screen. If an FPO manager asks, “How much soybean did we collect today, and what payments are pending?”, the system should answer exactly that, without three menus and a filter.
- Test for dialects and real field conditions. “Supports Marathi” means little if the model fails on Varhadi in Vidarbha or Ahirani in Khandesh. Test with native speakers on noisy mandi calls, weak connections and older handsets. Check crop names, local measures like maund or guntha, and amounts, and confirm high-stakes values by voice before any transaction completes.
- Keep a human in the loop. AI should know when it does not know. When confidence is low, or when credit, payments or disputes are involved, it should hand over smoothly to a trusted person, such as an FPO coordinator or field agent.
- Treat voice data as a trust contract. Voice recordings are personal data. Collect them with informed consent in the user’s own language, use them only for the stated purpose, and comply with the Digital Personal Data Protection Act from day one.
- Count who is missing. Track adoption and success rates by gender, language, age and landholding size. If women or tribal farmers in a district use the system far less than others, treat it as a product defect and fix it with the urgency of a production outage.
The first wave of agritech digitised transactions. The next must democratise them. When a woman farmer in Vidarbha can speak to technology the way she speaks to a trusted neighbour, and get an answer she can act on, AI will truly belong to Bharat.
Express Computer is one of India's most respected IT media brands and has been in publication for 24 years running. We cover enterprise technology in all its flavours, including processors, storage, networking, wireless, business applications, cloud computing, analytics, green initiatives and anything that can help companies make the most of their ICT investments. Additionally, we also report on the fast emerging realm of eGovernance in India.