Why one cow tech company has already integrated ChatGPT into its product line up
Article originally published in the Cow Tech Report
June 19, 2023 by Walt Cooley
One of the hottest topics in tech right now is artificial intelligence. When I heard that a Dutch-based company had integrated an AI-powered feature into its software for dairy farms, I reached out to see what they had achieved with the cutting-edge technology. I also wanted to discover how easy it was to add an AI integration.
Connecterra software engineers and product developers recently explained to me why they have added a ChatGPT-powered suggestion prompt into their Ida Enterprise software’s Game Plan Tool for farms.
First of all, let’s get a few definitions out of the way. Skip ahead if you feel comfortable with this terminology already.
This new integration was developed during a hackathon. A hackathon is a sandbox for software developers. For a set period of time outside of their regular work, they get to play around with code and develop use cases without the pressure of building out a fully-fledged solution. Connecterra senior software engineer Alex Vyshnivskyi says it gives engineers like himself the opportunity to “try out new ways of working and thinking.” Creativity and innovation are the goal.
AI (artificial intelligence)
Artificial intelligence (AI), like ChatGPT, refers to the field of computer science that focuses on creating intelligent systems capable of performing tasks that typically require human intelligence. AI systems, such as ChatGPT, are designed to understand and generate human-like responses to questions, prompts and conversations. In short, it can automate tasks based on receiving text-based prompts from a human.
This company was founded in 2015 to help dairy farms increase productivity while reducing the impact of milk production on the planet. Its tag business was recently acquired by Datamars. Connecterra now focuses exclusively on its SaaS platform. Its Enterprise platform is powered by real-time farm data and displays farm data visually to reveal high-value insights and farm trends. The company’s Game Plan Tool allows farmers to set a farm-specific goal over a defined period of time, make assignments to team members to initiate change and have the software use the data available to it to track if the farm meets the goal.
So, for example, if a farm wants to improve rumination time the company’s software can help them set and achieve that goal.
Connecterra’s Game Plan Tool is where the company found a use case for ChatGPT.
Vyshnivskyi says the use case he and other engineers tested and have since adopted is a single push-button suggestion prompt that will give farmers proposed actions they can take to help them achieve their goals. Take for example the previously mentioned goal of increasing a herd’s rumination time. With the click of a button, farmers can now have a list of suggested actions to review and add them to a ‘game plan’ to achieve a goal such as improving rumination time.
See it in action here.
What’s going on in the background when a user clicks the “Recommend task” button is the initiation of a series of events that Vyshnivskyi helped to code. That code gives text-based instructions to ChatGPT and asks it to do research on best practices to help meet the farmer’s goal.
“Basically, it works like an assistant that helps you. You’re telling it: ‘I have this goal on my farm. I want to get to only a certain number of cases of ketosis or increase my milk production or so on.’ Then it gives recommendations on actions that you can take that actually help you to achieve your on-farm target,” says Nynke Slegten, director of product management for Connecterra.
The recommendations that it generates are dynamic based on the farm goals that users set. The integration harnesses ChatGPT’s capabilities to scour the internet for academic research, summarize it and suggest actions to take.
Vyshnivskyi says that while it was pretty to easy to integrate with ChatGPT’s application programming interface (API). The most challenging aspect of the hackathon was learning to become an AI prompt engineer and give the right context and parameters to have AI return a relevant response.
“It’s a bit different than other software engineering because you need to explain to the AI model how to behave,” Vyshnivskyi says. “For instance, you need to say to the AI, ‘Okay, act like a farm assistant. All of your knowledge is limited to the dairy industry. You are not allowed to answer any other questions besides dairy questions. It was challenging to fine-tune the prompt and make it really useful.”
Vyshnivskyi envisions software companies using AI integrations will probably have specific engineers who specialize in AI prompt engineering in the future.
“This technology integration is automating a lot of the research a farmer might do,” says Jenn Gbur, director of product marketing for Connecterra. “In the past, you might have done a little digging on some of the academic papers around the topic of say, ketosis, if you’re having that issue on your farm. Or you might have picked up the phone and called your nutritionist or your veterinarian and asked for advice. Tools like ChatGPT can automate some of this work. It can give you a starting point to help you have a more informed conversation when you pick up the phone and call your nutritionist or veterinarian.”
The company clearly notes that no personal farm information is being shared with ChatGPT through this integration.
“The hackathon was a great way for us to try out multiple application integrations of ChatGPT within our product,” Slegten says.
Since the hackathon, the company has developed another integration with AI. This one can give a farmer advice with the single click of a button based on the data trends their farm data dashboard is displaying. Again, no actual farmer data is being communicated to the AI. The software has been engineered to translate the data trend into text based instructions that ask AI to suggest recommended actions.
So, for example, if DMI increased and refusals decreased, a farmer using the company’s software could ask AI to suggest reasons why that may have happened.
See this feature in action below.
“With this feature, what you’re effectively doing is running through the list of the easy things you can do and then if it’s still not resolved, you’re using the time that you have with your nutritionist in a more efficient way to problem-solve,” Gbur says.
Connecterra says that early feedback from producers who have tried these features shows they like it. These users like that it saves them time.
“Normally a producer would go to Google and browse through academic articles, but to go through those yourself takes more time. But to have it summarized right in the software when you are doing the analysis can broaden perspectives, remove biases and save you time getting to a solution,” Slegten says.
Connecterra sees more integrations with AI going forward.
“We’ve been using different types of AI in our products since our founding over eight years ago. But I think just the speed and the evolution that we’re seeing with generative AI and other AI technologies is truly mind boggling,” Gbur says. “The way that AI evolves so that it helps us help other people is going to continue to change. It’s an exciting time. It’s a really fast-moving time, but it’s super cool.”
“What is very interesting about AI is that you now have access to large amounts of data, volumes of data, and knowledge bases that get combined and translated into something that is human and you can just ask a question to. Imagine you have this very powerful employee that is learning extensively all the time.” —Nynke Slegten, director of product management, Connecterra