On a 60-cow registered Holstein dairy outside Baldwin, Wis., artificial intelligence has become part of the management toolbox.
Mary Holle and her husband, Joe, milk in a refurbished 94-cow tie-stall barn at Holle-Oaks Dairy, a family operation that has seen major change in recent years. After taking over the farm from Joe’s parents in 2017, the Holles made a hard pivot in 2024, downsizing from 120 cows to 60.
With her father-in-law ready to step back from daily chores and labor costs continuing to climb, Holle could see the pressure building. The farm had reached a point where something had to change to keep things sustainable for everyone involved. Thus, downsizing the herd became the path forward.
Part of that downsizing process involved bringing in a tool still new to many dairy farmers — artificial intelligence, or AI. It wasn’t an obvious fit, but Holle saw it as a way to work through her farm’s numbers and run different scenarios without adding more layers to an already full system.
A Numbers Mindset Meets a New Tool
Holle, who also serves as the program manager for the Farm and Industry Short Course at UW-River Falls, didn’t come to AI without experience. She’s long leaned into data, building her own systems to track and understand how her farm is performing.
“I’ve been using Excel since I was, like, 10 years old,” Holle says. “I started doing my dad’s dairy herd records, because we didn’t milk test.”
That early passion for data turned into formal training in dairy science and ag business, along with several years of building detailed spreadsheets for her own operation. Today, those workbooks track just about everything on the farm, from feed costs and veterinary expenses to crop yields, soil tests and labor.
“I’ve been building complex equations within Excel for like a decade,” Holle says. “My biggest workbook is 17 pages long.”
Even with all of that in place, she eventually hit a point where spreadsheets alone weren’t enough to work through the number of what-if scenarios she was running. She wasn’t trying to replace the system she already had, but she needed a faster way to test ideas and see how different decisions might actually play out.
That’s when Holle began to use AI. She started experimenting with it by feeding in her own farm numbers, then asking it to run different scenarios and compare outcomes she would normally have to build out by hand. Over time, she used it to work through decisions faster and feel more confident in what the numbers were pointing to.
Running the Numbers
The 2024 transition pushed Holle to take a closer look at her cost structure. With fewer cows, fewer employees and new financial obligations, she needed to figure out what her cost per cow and break-even milk price needed to be for the smaller herd to stay profitable.
“I needed to run a series of scenarios to find the linchpins in the business,” Holle says. “We had to drop our cost per cow and get our break-even down to around $17.80 per cwt. for the smaller herd to work.”
She pulled data from her workbooks, including fixed costs, five-year averages for feed and vet expenses, labor hours, wages and loan balances with payment schedules. From there, she used AI to organize the information and get a better read on what was driving cost per cow.
“I asked ChatGPT, ‘What are the trends, what’s going on, can you put this into context?’” she says.
Working in short windows between chores, Holle ran different scenarios around debt, labor and herd costs to see which changes would have the biggest impact. It didn’t hand her one answer, but it helped narrow the decisions down.
“It told me what needed paid off first and where I’d see the most return,” she says. “I took the results to our banker and he said, ‘That’s ingenious.’”
In the end, the Holles were still the ones giving the final say, but AI helped them sort through information quicker and feel better about the direction they were headed.
Taking AI to the Field
After using AI to work through the financial side of the operation and guide the downsizing decision, Holle started looking at where else it could fit. Crop management was the next place she turned, and it’s something she’s still working through this year.
The farm currently includes about 500 acres in rotation with corn, soybeans, alfalfa and wheat. Similar to her herd management Excel work, Holle had built up soil tests, yield maps and field histories over time, but the information wasn’t connected in a way that made it easy to use. This year, she started using AI to organize it by field and year, then layer in crop history and yield data so it could be compared more directly.
“I didn’t have everything tied together in one place,” she says. “I had the information, it just wasn’t organized in a way I could actually use.”
With a new structure in place, Holle began asking AI more targeted questions around nutrient management and input efficiency. One focus centered around nitrogen — how much was already available in the field and where she might be able to cut back on applications without hurting yield.
“I wanted to know what was already out there before just putting more on,” she says. “If there was a place to save a little money without giving up yield, I wanted to find it.”
She’s also started looking at whether past decisions, like planting BMR corn, may have longer-term effects on nutrient availability. Using AI helped Holle spot patterns and show up to conversations with her agronomists better prepared.
This spring, she and her agronomists used that analysis as a starting point to fine-tune fertilizer and spray programs by field, paying closer attention to residual nutrients and timing. The new plan cut back on total fertilizer and chemical use compared to the previous year. By her estimate, this adjustment will trim roughly $40,000 from her fertilizer and spray bill in 2026.
Keeping Perspective in Place
While Holle sees value in AI, she’s careful about how she uses it. Sensitive information stays out, including personal identifiers, financial accounts and tax data.
“There’s always a step between sensitive information and it,” she says. “Anything personal or financial doesn’t go straight in. It always gets filtered or kept separate first.”
That caution carries into how she uses the tool in day-to-day decisions. Even when AI is helping her work through parts of the farm’s data, it hasn’t taken over decision-making. Holle still relies on her own judgment when something doesn’t line up with what she’s seeing on the farm, especially when context doesn’t show up in the numbers.
“Yes, I think it’s made me a better farmer,” she says. “But it’s a tool for the areas where I don’t know enough. There’s always context it’s going to miss. You can read a person or a situation in ways it can’t.”
For other producers thinking about trying AI, Holle recommends starting small and treating it like any other tool on the farm. Don’t start with big decisions or sensitive financial work. Start with something simple, learn how it responds and build from there.
“Start with emails or documents,” she says. “Something low risk where you can see how it responds and get comfortable with how it handles your information before moving into anything bigger or more complex.”
From her experience, it has less to do with the technology itself and more to do with how organized the farm’s information is going in. If the inputs are messy or incomplete, the results will be, too.
“Garbage information in leads to garbage answers out,” Holle says. “If you don’t know what you’re asking for, you won’t get what you need.”
That also means knowing where the farm stands before expecting any tool to improve it. Clear records, numbers and a good handle on what’s working and what isn’t all matter just as much as the software being used.
“You need to understand your strengths and weaknesses first,” she says. “Know what you’re comfortable handling on your own and where you could use a little more support, so you’re not leaning on the tool for things you already do well or expecting it to fix gaps you haven’t identified yet.”
Faster Decisions, Tighter Management
On her 60-cow dairy, AI hasn’t replaced hands-on management or day-to-day decision-making. Instead, it’s helped her sort through financial choices, tighten input decisions and show up to conversations with advisers with more clarity around the numbers.
For Holle, it’s become a fast, free tool she can pull up anytime to work through questions and run scenarios. And it’s helped her move through decisions faster and keep the operation running a little tighter, without adding more layers to the process.
