It was only ever a matter of time before ChatGPT came for agriculture.
Generative AI has already infiltrated restaurants, ingredient development and other areas of agrifoodtech. Now, it’s providing agronomy advice to growers courtesy of Norm, the recently unveiled AI chatbot developed by Farmers Business Network (FBN).
Farmers registered on FBN can access Norm for free (for now) to access near-real-time information about farm equipment, planting times, soil health, and many other areas of farming.
A ‘significantly different technology’
FBN’s offerings address multiple parts of the farm operation, from crop insurance and equipment loans/mortgages to seeds and livestock medicine, not to mention finance.
“There are so many different areas where we think we can help farmers maximize ROI,” Kit Barron, head of Data Science at FBN, tells AFN.
Using technology to assist with this isn’t new for FBN. Barron says the company already relies on some traditional software to help with its agronomic advisor roadmap. However, “when ChatGPT came out it provided us a base solution that can help us efficiently solve a lot of those different use cases, in a much quicker-to-market pace.”
“This is such a significantly different technology that it’s allowed us to iterate really quickly add capabilities in a lot of those areas.”
What Norm is
Norm — still in beta mode and only available to those in FBN’s network — is built on OpenAI’s Chatbot GPT-3.5. To answer queries, it pulls from publicly available data on weather, soil monitoring, application rates, current events and other areas relevant to farmers. It also uses FBN’s own proprietary data feeds, which are specific to agronomy, animal health, crop health and the overall business of farming.
The name Norm is a hat-tip to agronomist and Nobel Peace Prize Laureate Dr. Norman E. Borlaug, who is often called the “father” of the Green Revolution. Borlaug’s work breeding high-yield, disease-resistant crops in the mid-20th century, drastically improved crop yields and food security.
Norm the chatbot probably won’t be winning a Nobel Prize anytime soon, but its capabilities can assist farmers in FBN’s network with a range of questions. FBN says some areas where Norm especially excels are chemical intelligence, input guidance, pest and disease strategies, livestock and animal health, and “general agronomic advice.”
As it’s planting season in the US, Barron says many of the questions submitted so far are about soil conditions, soil temperature, the right time to plant.
“We’re also seeing a lot of questions around crop protection, particularly for pre-planting use cases. [There are] a lot of really interesting questions on types of tillage equipment.”
He adds that FBN envisions Norm as “a first-line advisor to our members across dozens of topics.” (FBN stresses that Norm is not the final word on agronomy advice, however. More below.)
Farmer input from the start
Growers and farmers have been part of the development and feedback process from the start, says Barron. “We worked with our internal agronomists and veterinary medicine team and animal health team to ensure that the quality of the answers that Norm was providing was consistently high.”
“We had internal testers and then we opened it up to 50 to 60 of our most-engaged members that we’ve periodically reached out to for things like this and got their initial feedback.”
He admits the team was a little nervous initially, since farmers aren’t necessarily early adopters when it comes to tech.
“But people were overwhelmingly positive and saw the potential of this. That kind of gave us the green light to get this beta out there.”
What Norm isn’t (yet)
Barron stresses that Norm is not a replacement for agronomy advice. “We don’t want to mess around when it comes to to animal health or crop protection.”
Agriculture in the US is governed by a complex web of legal and regulatory requirements that can vary from one state to the next. There are also legal concerns around things like recommending controlled-use pesticides on a farm, as mis-use could destroy an entire crop and be a threat to human and animal health.
Barron says all of this is “top of mind” for FBN, which already follows strict protocols across the rest of its business.
For example, in its e-commerce store, FBN must know whether a product is labeled for sale in every state or not, and sell to farmers accordingly.
“We’ve spent a lot of time structuring data around the what we call product labels, which tell a farmer what a particular product is or isn’t registered for,” says Barron. He says these aren’t so much product labels as enormous PDF files that “can be dozens of pages long.”
“The EPA has to approve every single combination: ‘you can use this product in this state under these conditions at this rate mixed with this other product.’ It’s incredibly complicated.”
That alone accounts for the warning affixed to each of Norm’s answers: “Norm is an experiment: Not intended for real agronomic guidance. Always follow label directions and local regulations.”
FBN is working to integrate more of this complex data into Norm by leveraging the company’s already existing data assets.
“We’ve been able to point Norm at that structured dataset. That’s kind of been our approach to ensure [accuracy] around some of these more sensitive uses. Our growers are getting really high quality responses.”
But, he cautions, Norm is still a beta. “We do expect there to be mistakes, and we want to learn about those as well. And so we’re being really clear that this is this is an experiment. This is not intended for real agronomic guidance at this time. It’s really important to consult with your local regulations and the product label and your trusted agronomist or FBN advisor. That’s how we protect ourselves and protect our members and set the right expectations for what this product can and can’t do today.”
However, there’s plenty of opportunity for Norm in the ag world, he adds. “The conversational form factor is so consistent with how so many farmers already interact with their advisors. A surprising amount of business is done through SMS in agriculture. This chat interface could become a principal way to interact with a lot of parts of the FBN experience.”
Guest article: Farmers aren’t buying today’s ag robotics model. You shouldn’t either