AI: a new assistant

MADISON, Wis. — Imagine being able to obtain a herd performance analysis, troubleshoot equipment problems or translate documents for multilingual staff within seconds. Envision the possibility of forecasting a drop in production or detecting mastitis before it occurs.

These are examples of how artificial intelligence can assist today’s dairy farmer.

During his presentation, “AI on the farm: Smarter herd management,” March 5 at the Professional Dairy Producers Business Conference in Madison, Jeffrey Bewley shared how farmers can benefit from AI.

Bewley is the executive director of genetic programs and innovation at Holstein Association USA Inc.

“These are exciting times for those who love both data and cows,” Bewley said. “AI is beginning to revolutionize the world around us and the dairy industry. AI is the biggest breakthrough for the industry since TMR (total mixed ration). We have so much data coming into our operations, and we can use AI to help us make better decisions. There is a lot we can do to improve profitability of our farm by looking at data.”

Bewley defined AI as machines that learn from data and make predictions or decisions.

“AI is not science fiction,” Bewley said. “It’s math, statistics and computing power. AI is computers mimicking human intelligence.”

Today’s AI is faster, cheaper and more accessible than ever and advancing rapidly with new releases coming out every 2-3 months, Bewley said. Large language models like ChatGPT and Claude are two popular versions.

“You don’t have to know anything about computers to use a LLM,” Bewley said. “You just talk to it, like you talk to another person. It learns our habits and who we are. It’s mindboggling what these tools can do.”

Bewley said the most successfully adopted form of AI in the industry is wearable activity monitoring sensors.

“This transformed the way we manage cows,” he said. “Many of these sensors employ AI to do the predictions for temperature, activity, ruminating and eating.”

AI can be found in other products as well. A product using AI to manage people in the parlor identifies how well they adhere to procedures. A precision soaking system that uses AI reduces water usage by about 70% by delivering water if a cow is standing where soakers are located along the feed bunk. An automated calf feeder with AI capabilities provides information on demand regarding the drinking status of individual calves.

Bewley said the industry is heading toward technology that will allow data to be accessed by typing in or speaking into their phone and saying, for example, “I want to know my pregnancy rate for first-lactation cows for the last year,” or “I want a list of all cows below 30 pounds of milk.”

AI can be used to improve feed efficiency by optimizing nutrition on a per-cow basis and improve animal welfare through continuous comfort assessment. AI can also be used to detect digital dermatitis, measure feed intake and provide process automation.

AI can help manage risk in milk and feed prices by forecasting future price trends based on historical data and market conditions, optimize purchase and sales strategies to maximize profit, simulate market scenarios to plan for price volatility, and help manage feed inventory efficiently to reduce waste and costs.

To demonstrate how machine learning works, Bewley provided an example for lameness detection.
The first step would be to collect the data. In this case, that involved 10,000 videos of cows walking, and the data was labeled “lame” or “not lame.”

“AI doesn’t know by just looking at those videos if a cow is lame or not,” Bewley said. “Somebody has to say, ‘These are the 100 cows that were lame, and these 9,000 were not.’ You have to train it, but that only has to be done once.”

The algorithm learns gait patterns associated with lameness and then flags lame cows from footage.

Predictive analytics can help farmers make decisions about reproduction, mastitis management, ration responses and more. For example, when looking at lactation curves, AI can help predict a cow’s production level at dry off to determine the best time to breed her.

AI predicts problems before they become visible.

“Predictive analytics is where we’re going to see a big change in how we look at data,” Bewley said. “How can I predict what my risks are for mastitis? What are my risks for a metabolic disease outbreak or drop in production? AI can help us understand all these complex relationships in a way traditional statistics can’t really do.”

Bewley once uploaded a herd summary into ChatGPT and told it to analyze the herd. He asked it to identify key performance indicators and provide strengths and weaknesses in a dashboard format. Within seconds, without Bewley putting in any additional feedback, it read the herd report and provided a report.

“I know this herd, and I would have identified most of the same weaknesses,” he said. “It was a pretty good report.”

AI can summarize research papers in plain language, draft employee standard operating procedures and training materials, analyze Dairy Herd Improvement data and explain trends, generate interview questions for hiring, create social media content, explain nutrition concepts, help with regulatory compliance questions and brainstorm solutions to management challenges.

To get better results, Bewley said it is important to ask the right questions.

“Be specific and intentional,” he said. “A vague prompt would be, ‘Tell me about mastitis.’ A specific prompt would be, ‘I have a 500-cow Holstein herd on sand bedding with rising SCC (somatic cell count). Create a mastitis prevention protocol for fresh cows.’”

Other example prompts for farmers include: “Here is my DHI data. Identify the three biggest opportunities for improvement and suggest specific actions.” “Create a step-by-step training guide for new milking parlor employees. Include safety protocols and common mistakes.” “My TMR mixer is producing inconsistent batches. Walk me through a diagnostic checklist to identify the problem.”

Voice-activated farm management is an example of AI-based products in development for use in the industry. This would enable a farmer to ask their phone, “How are my fresh cows doing today?” Another would be predictive health alerts in which AI spots illness 5-7 days before symptoms appear. Expanded robotics that are more affordable and versatile are also in development.

Bewley said to use LLMs for brainstorming and drafts but verify facts independently.

“Don’t trust AI over your own observation, and don’t assume it’s always right,” he said. “Use common sense when looking at the data.”

From troubleshooting problems to finding improvement opportunities, farmers can leverage the power of AI to achieve greater success in their operations.

“AI isn’t coming to dairy; it’s already here,” Bewley said. “The question is: Will you use it to your advantage?”