‘What You See With KissanAI Is Kind Of The Ultimate Learning': Founder Pratik Desai On How This AI Chatbot Is Helping India's Farmers

In less than a year, artificial intelligence (AI) technologies have risen rapidly, spreading their influence over almost every aspect of human life. While many have advised caution, it is inevitable that AI will alter several sectors in ways both good and bad. Striking while the iron is hot, an AI platform is trying to revolutionise agriculture in the world's second largest food producing country.

KissanAI, a multilingual AI chatbot, was launched in March this year. The platform is engineered to provide farmers with personalised, voice-based assistance for all their agricultural needs, from specific pest control for their location and crop to what is the right time to plant a crop for the best prices. The platform has already had over 100,000 users so far. It leverages expert information, gathered from researchers and agricultural universities to provide information to farmers that they may not find on the internet with a quick Google search.

"Everything is done through AI. So there is no person behind who is providing biased advice… we are actually getting this information from recommendations that have been generated by universities and professors," Pratik Desai, founder of KissanAI told The Core.

The platform could address significant challenges faced by Indian farmers, like market price fluctuations and climate change-driven weather inconsistencies. "When you are relying on your neighbour, who is not an expert in climate, for more information there is going to be a lot of misinformation and a lot of conspiracies. But with AI, it would be factual," Desai said.

To understand how the platform could be scaled up, how sustainable is the business of AI in rural India and if this paves the way for more AI penetration in India, we caught up with Desai.

Here are edited excerpts from the interview:

How did you come to start KissanAI and what made you choose agriculture to work in?

My PhD, which I completed in the US, was focused on machine learning (ML), artificial intelligence, semantic web technology, mostly on the knowledge extraction side and after that, I moved to Silicon Valley where I founded three different startups. In 2019, I decided that I want to focus on Indian agriculture, bringing AI and ML applications.

It was obvious for me because my family has been in agriculture for generations - my father is still a farmer. And I have been associated with agriculture myself, working in the farmland before I moved to the US, so I was aware of all the problems they face and what challenges Indian farmers have. While I was here , I also saw how data logistics and other information technology and ML approaches are helping farmers in the western elevate their way of living. With my background in this field, I thought that there is a big opportunity for me to bring the knowledge and experience that I have to this domain and find the unique challenges, identify them and create a startup out of it.

What you see with KissanAI is kind of the ultimate learning that we have achieved after four years developing different applications. We learned their (farmers') habits, what they're actually looking for, what are the knowledge gaps, information gaps…and with the feedback that we're getting, we think we've found the nerve of the whole agriculture ecosystem and we can actually be very useful and ultimately help the people on the ground in bridging this knowledge divide.

Could you explain what concerns KissanAI addresses, and also your other platform Titodi, which is also in the agriculture space?

So Titodi is actually the name of our parent company. Titodi is actually a farming bird… when I was on the ground working with my dad I heard a lot of stories about how it's a farmer's friend. So we decided on that name because we want to be the digital farmers' friend. We developed four different products before KissanAI. What you see as the Titodi platform now is the marketplace, which we started in May 2022, where farmers can have their independent online running store. We have more than 400 stores by now by farmers, some FPOs and other retailers. While we were building this, we were collecting huge amounts of agriculture knowledge base. Something we realised is that if a farmer is not educated, they have language barriers, they have problems with writing and coming up with marketing related thoughts. So if I ask a farmer to list something on Facebook marketplace, they will not come up with copywriting words, so we help farmers generate content when they're trying to sell something.

And this was the basis of what you see in KissanAI today. We have been collecting huge amounts of information that we can now use to generate the insights that you see on KissanAI. Now, the first concern was a language barrier. The other was literacy, and Google searching everything, which is complicated. We also saw in many cases that farmers would ultimately rely on the person next door who may be a little more expert than them… so there's a herd kind of thinking. That's why we created something where the solution is bridging the knowledge divide.

Everything is done through AI. So there is no person behind who is providing biased advice because sometimes farmers also rely on agri companies for inputs, or when you call someone local or a shopkeeper, he has a bias towards selling his own items. Here's also where we're helpful because we are actually getting this information from the recommendation that has been generated by universities and professors.

How many users does KissanAI have and where are they located geographically?

We have users from all parts of India, we're serving ten languages. It has picked up through online influentials - people have been tweeting about it, people has been making reels and videos, and so on. We have had more than 100,000 users try it out so far.

You mentioned how it gives unbiased information and is using expert knowledge. So, where is the information coming from for KissanAI?

We use ChatGPT as a base. From the first week only, OpenAI has been supporting us with a grant and also technically. Microsoft is doing the same, it has given us a good amount of Azure credit so we can host our server there. But when you are training a large language model, you are taking all available information for every country, not just India. So, if you ask what the different varieties of banana are, it can be anything, the answer is not very specific.

So we have collected recommendations from agricultural universities, which are mostly not available on the internet. So whenever a request comes in, it comes first from our knowledge base which is India specific, and if it's not found it then we rely on ChatGPT. And we're also growing this knowledge base. We're working on collaborating with the Ministry of Agriculture and National Bank For Agriculture And Rural Development (NABARD) too. Our goal with them is to create a one-of-its-kind agriculture and marketplace where everything in India ever produced for agriculture is under the one repository.

How many universities are you working with currently?

Currently we have a memorandum of understanding (MoU) with one university, the Navsari Agricultural University in Gujarat, but we are working with 38 universities to collect information from.

What has the response from farmers and others who have used the platform been like?

We've seen a huge response once this , where the farmers have actually started coming back with feedback and suggestions. We're trying to build new features based on the feedback. One of the things is getting access to market data. For example, if someone produces something, they don't know which mandi will give them a good price. It's not very easy to access the prices - it is collected on government websites, but it's not very directly accessible to them (farmers).

Do you think this feature would be useful to farmers in times of inflation, for example right now, when tomato prices are skyrocketing?

I am not worried about a time when the prices go high. I'm worried about the time when the prices go very low. And this market price feature will be very exciting...it's going to help them to find out which months will probably have the lowest price and when you should not be growing, what else you should be focusing on, etc. We're enabling them to ask AI not just what was yesterday's price, but what was last year's price in February.

And in the future, we want to actually integrate weather information with KissanAI. Once the farmers start creating accounts, they could start actually getting regular weather updates and if it's going to affect the crop.

How do you plan to scale up in the near future?

So KissanAI is farmer facing, and we're not charging farmers. But we are also building something of an AI API platform, which will have our whole linguistic pipeline with the voice and text and the agriculture knowledge base, that agriculture companies which are serving users in like India, where the language is important, can integrate inside their application or website. Every farmer in our community have companies (fertilisers, seeds, etc) they mostly rely on and there's challenges for these companies on how to deliver knowledge about their products. The API platform is basically a business model where the KissanAI version of that particular company can deliver answers to their customers directly.

When it comes to farmers, what kind of challenges do you have in sort of making this accessible to them? Do you see technological, logistical issues as a big problem?

No, I don't. So this kind of application is not Instagram, or YouTube, it's like you ask a question once a day or once a month. And for that the internet and the devices that we have are sufficient enough, because it will not require a lot of bandwidth. Actually, I see how India's internet is cheap as the enabler instead of the . Internet and mobile phones are everywhere and its going to get more prevalent. So this is not going to be the challenge at all.

Right. And in terms of reaching people, making them aware that this platform exists, how is that happening? Is it happening through word of mouth or through some other channels?

It's all word of mouth right now. I'm not doing any marketing or publicity. Someone will like it and make a reel or video and share it. And in terms of awareness, I have seen that people who are in very privileged areas, they still have resistance… they may say why should I use AI, I know everything. Most of the people actually are coming from challenging areas like Chhattisgarh, Saurashtra, Vidarbha, areas of Karnataka, Andhra Pradesh, where they have to struggle with water and so on…they are actually the ones using it the most because they're always looking for solutions to get better.

Microsoft also recently announced a rural AI chatbot – Jugalbandi. Can we expect to see a lot more AI penetration in rural India across all sectors in the near future?

Jugalbandi is a policy chatbot. We are doing agriculture, we're also exploring something in the rural education space. But there are domains that are still a little bit not mature for AI, like health for example.

The biggest problem is sustainability. You can create something not as a startup, but as a foundation getting grants, because at this point, making money from rural India is very difficult for startups. You have to find a domain where someone actually will be using your technology to be integrated inside their platform. Otherwise, if you're creating a bot for rural India, AI is very expensive in terms of GPU and API charges, and so on. I see some coming up that are probably supported by government agencies or funded by them. But in terms of a product, I think it will take some time.

Climate change is something that is increasingly affecting Indian weather and weather patterns, and you can see it impacting agriculture. So, do you think AI and KissanAI can address this and help farmers navigate what is happening?

Yes – what we're trying to do is create a one-stop assistant and build trust with farmers. When you are relying on your neighbour, who is not an expert in climate, for more information there is going to be a lot of misinformation and a lot of conspiracies. But with AI, it would be factual, and if we have established a relationship with the farmer, where they rely on us, this is the relationship we are trying to build.

Then all the climate related knowledge gaps or facts or new techniques, we can deliver to the farmer. It can help them with sustainable farming, climate change.

There are many non-profits doing work in remote areas, that we are working with, with a focus on sustainable agriculture. What we want to do is be the one-stop platform for authorities to pass the message, and the same platform will deliver the message to everyone on the ground.