Reworking the economics of farming, with OpenAI in the loop

Most Indian farmers know how much crop they produce. Traders know market prices. But very few know what they actually earn per acre. Costs are rarely tracked in detail. Selling decisions are often made without visibility into margins. Over time, that lack of clarity compounds into shrinking profits.

Sachin Farfad Patil grew up watching this play out in his own home. His father farmed in Maharashtra, growing soybean, cotton, chickpeas, and wheat. Like many farmers around him, he sold produce without fully accounting for cost of cultivation, current market prices, or even the government’s minimum support price.

“They were following the same practices for decades,” Sachin says. “But profits per acre kept shrinking.”

After 14 years in corporate roles, he returned to that question. Two years ago, he started building GramIQ.

From information to decision-making

GramIQ focuses on a basic problem. Farmers don’t lack information. They lack usable, contextual intelligence.

The platform brings together three layers that typically exist in isolation: financial tracking, crop advisory, and access to real-time information such as mandi prices, weather, and government schemes. On their own, each solves part of the problem. Combined, they begin to influence decisions.

A cotton farmer checking prices might see that the current mandi rate is below both the minimum support price and his cost of cultivation. Instead of selling immediately, he can wait. That shift from reactive selling to informed timing is where the value begins to show.

Watch now

Building across crops

For crops like cotton, rice, and chili, GramIQ has built proprietary models with agronomic expertise embedded into them. But Indian agriculture spans dozens of crops, each with its own variables. Building custom models for all of them would take years.

For crops where such models do not exist, the system relies on OpenAI models such as GPT-4o and GPT-4 Turbo to interpret farmer queries and generate responses in local languages.

Accuracy has improved from around 65–70% two years ago to 85–90% today. But AI does not operate alone. Every response in crop advisory is reviewed by an agronomist before it reaches the farmer. Sometimes it is approved. Sometimes it is adjusted based on local conditions, availability of inputs, or weather patterns.

“We’re dealing with livelihoods,” Sachin says. “AI helps us move faster. The expert ensures we don’t get it wrong.”

Scaling across languages

Agriculture in India is deeply local. Language, practices, and terminology vary across regions. Scaling requires systems that can adapt without being rebuilt each time.

GramIQ currently operates in Hindi, Marathi, Gujarati, and Bengali, with more languages being added over time. OpenAI’s multilingual capabilities allow the platform to expand without re-engineering the entire system for each state.

“We need to translate conversations, advisories, and insights into different languages,” Sachin says. “AI helps us do that without rebuilding from scratch.”

Handling queries at scale

Today, GramIQ serves over 1.2 lakh farmers, with about 60–65% actively using the platform each month. That translates into 400,000 to 600,000 interactions every month.

A large share of these queries, especially for crops without proprietary models, are processed through OpenAI-powered systems.

“Farmers don’t ask in structured ways,” Sachin says. “They use mixed language, local references, incomplete sentences. The system still needs to understand and respond correctly. Building that from scratch would have been extremely difficult.”

The conversational layer itself was prototyped and launched in under two months. Without OpenAI handling natural language processing, the same effort would have taken significantly longer with a larger team.

When AI reaches the field

In the early days, users assumed responses were coming from a human team. One of Sachin’s relatives even called to ask who was replying from the backend.

That gap is beginning to close. Over the past few months, GramIQ has trained thousands of farmers through hands-on sessions designed to familiarise them with AI-driven tools. As usage grows, so does trust.

Early data from the platform shows measurable shifts. Cost of cultivation is down by around 15%, while profits per acre have increased by nearly 20%. These changes come from better tracking, more informed selling decisions, and timely advisory.

From farmers to farm-preneurs

The plan is to reach 20 million farmers over the next few years. At that scale, decision-making across agriculture could begin to shift from guesswork to data.

The implications extend beyond individual farmers. Aggregated insights could inform how banks lend, how agri-input companies operate, and how government programmes are designed.

For Sachin, the shift is ultimately about mindset. “Farming becomes less uncertain when you can see the numbers,” he says. “When you understand your costs and your margins, you start making decisions differently. We want farmers to become farm-preneurs.”

The future of agriculture, he believes, is not just about better inputs or machinery. It is about better decisions. OpenAI did not create that vision. It made building toward it faster, and at scale, possible.

Watch now