The Moos Room™

Dairy farms generate plenty of data—but turning those records into useful management decisions takes time. In this episode of The Moos Room, Brad explores how artificial intelligence can help organize dairy records, summarize herd performance, and identify priorities for follow-up.

Brad shares his experience using Claude AI and open-source dairy data tools to analyze DairyComp 305 exports from the University of Minnesota’s organic and conventional research herds. From milk production and components to fresh cow health, somatic cell counts, and culling, he discusses how AI can bring information together into a practical herd review.

He also describes using monthly DHI reports and milk fatty acid profiles to build summaries, graphs, and cow lists that make trends easier to spot.

Topics include:
  • Turning DairyComp records into herd summaries and suggested action items.
  • Tracking udder health, fresh cow infections, and milk component trends.
  • Exploring tools for ration formulation, feed costs, and lactation projections.
  • Combining multiple DHI reports into one useful management overview.
  • Recognizing the importance of complete records, data privacy, and checking AI-generated results.
The opportunity is making dairy data easier to use. Brad discusses how these tools could support farmers, veterinarians, nutritionists, and Extension educators as they work together to improve herd management.

https://dairydatascience.com/

https://github.com/padge01/dairy-data-tools/blob/main/README.md


Questions, comments, scathing rebuttals? -> themoosroom@umn.edu or call 612-624-3610 and leave us a message!

Linkedin -> The Moos Room
Twitter -> @UMNmoosroom and @UMNFarmSafety
Facebook -> @UMNDairy
YouTube -> UMN Beef and Dairy and UMN Farm Safety and Health
Instagram -> @UMNWCROCDairy
Extension Website
AgriAmerica Podcast Directory 

What is The Moos Room™?

Hosted by members of the University of Minnesota Extension Beef and Dairy Teams, The Moos Room discusses relevant topics to help beef and dairy producers be more successful. The information is evidence-based and presented as an informal conversation between the hosts and guests.

 Welcome to the Moose Room. It's November ninth and still away, and today I'm speaking in Pennsylvania about some of our calf studies in Harrisburg, and a little bit about fly control and some of the studies that we've done here at our research center at a organic conference in, in Pennsylvania. So I thought I'd go out and talk with some of the Pennsylvania dairy producers and share some of the things that we've done and learn a little bit about some of the challenges that they're having.

So hopefully it'll be a good time. But we'll talk about Sweden next week and some of the things on the Swedish dairy industry. But I wanted to tell you a- about something pretty cool that I saw in the last week, and it was on LinkedIn, and it was a post done by Padraig Lucy. So a Irish veterinarian living in California, somewhere in Central Valley, Bakersfield area, and he released things on dairy data tools.

So basically it's an open source add-on that lets AI read your dairies, Dairy Comp three hundred and five and EZfeed, which is a feed program from Valley Ag Software exports. And basically it adds the farm's export files to a folder, and then you ask plain words, "How is pen doing?" You know, "How is pen one doing?

How is pen two doing? What's in its ration? What's happening to a specific cow? What can we check on for the next visit?" So this is used for nutrition consultants or veterinarians that might be going to farms and trying to figure out what's going on and read all of the data that's coming off of these farms.

So basically these answers will come back as card summaries for cows, pens, diets, and basically it kind of helps you determine what's going on on a dairy. So I decided to give this a try. So it works on Claude AI, so in one of the AI platforms, and basically it helped guide me through which dairy comp commands to run, and it works with test day data or daily milk, and it reads events from different herds, and it uses the AI tools that he's built to basically give you a summary about what's happening.

So it tells you exactly which commands to run in Dairy Comp once you set this all up. If you have Claude AI, you, uh, put this little program that he has into it, and it tells you which programs to run or which, uh, Dairy Comp commands. You know, it gives you lactation number, pen, fresh dates. It keeps the sold and the dead animals.

Uh, it also looks for cow events from the last three years, and you can put EZfeed if you have EZfeed. We don't at our dairy, so we-- I don't have that. I tried to put some of the feed supervisor stuff in there, and it maybe kind of figured it out a little bit, but it's kind of fascinating what it can tell you.

You know, there's just a lot of data on these farms, and it gives you an idea about what's going on if you're able to record it. So let's talk about some of the things that I've run. So basically, I used our dairy. Now, I've only had Dairy Comp on our herd for about nine months or so. We're running dual programs.

I'm curious about what's happening with each. I think there's benefits to each one, and, uh, this one is Dairy Comp. And so I've put the data from Dairy Comp, and it's a little... You know, it doesn't have a lot of the historical stuff, like, and it's giving me a, a few things. But if you have, uh, complete Dairy Comp for quite a few years, it's probably a little bit better.

But it gives me an idea about what's going on. And it puts us in this wonderful output that you can look at, a visit review, basically. So a consultant, a veterinarian, extension, farmers can run this anytime, and it gives you an idea about what's going on each of the herds, uh, for that particular day, uh, with all of the data that you've provided it.

So I've looked at our herds in Claude here. I've taken all the dairy comp data, put it in there, and see what I could find. Well, it gives me, you know, how many milking cows, how many dry cows we has, how many cows have left within the last 12 months. Our median days in milk is 166. Our energy-corrected milk is 57 pounds.

Basically, we have two pens. We have our organic cows and our conventional cows. We have about 50 pounds of milk in the organic herd, 60 pounds in the conventional herd. Gives us an idea on mean days in milk. If you're recording a lot of the health data, it gives you an incidence of health issues within the first 30 days in milk.

So we're seeing about a 5.8% clinical mastitis rate and 1.5% metritis rate in the first 30 days in milk. You know, if I don't record that health data, obviously you're not gonna have that kind of information, so it's best to record all of that. You know, I try to record everything to see what we can get.

We've had 17 animals leave by 30 days in milk. Four have died. Um, 25 animals left by 60 days in milk, and so it gives you lots of culling statistics as well. It puts in the components, so we can see what our fat components are and protein for the last test. We can compare it with first lactation cows and older cows.

Now I'm telling you this, this has basically been reported out by this AI tool, and so it gives you this nice report where you can see that. Gives us an udder health somatic cell count over two hundred thousand based on the last test. So in our organic herd, we have about forty percent of the cows are over two hundred thousand.

Twenty-four percent in our conventional herd. Talks about repro. Now, the repro is a little spotty just because of our dairy comp, uh, not having some of the data. It wasn't able to recognize the bred events, so I've been kind of playing with that, trying to figure out what's going on with that. But it can give you a twenty-one day preg rate, uh, in the cows, uh, as well.

Now, the thing that I like the most about this, it kinda gives you some action items for a next visit or follow-up with the farmer or veterinarian, nutritionist, you name it, what you can look at. And it basically... So these are the action items. It tells us to walk both of the parlors. So we wanna look at teat end conditions.

We wanna look at our prep routine, post-dip coverage, and maybe milking forty cows with high somatic cell count at the end. It says we should culture our clinical cases to find the bug before actually choosing a control step. It says we should check our bedding, stocking, and cleanliness in pen one, that's the organic herd, because thirty-one cows were over two hundred thousand somatic cell counts.

It tells me I should find out what changed in August or September because clinical mastitis went up thirty per hundred cows in milk. So whether that's weather, bedding, milking staff, equipment. It tells us that our organic herd is averaging four point eight MUNs on its first test, so maybe we need to look at the nutritionist and see what's happening there on the MUNs.

It says we should decide which chronic cows to cull, either dry treat them, milk them last And then really look at some of these other health, uh, issues, DAs, ketosis, and see what's, what's going on in treating those fresh cows. So it's kind of interesting that you can see what is going on in a dairy farm just by putting it into AI tools.

Now, the other cool thing that Padraic has put out recently is a, basically a ration formulation software in AI. So it's available in AI, and you can run your own ration software, and it is quite fascinating when you actually look at this. It gives you, um, some ideas on example farms about what your income over feed cost is.

Uh, you can look at the different pens to see what's going on with what's in the ration, how much forage, how much protein. It gives you cost per day on all of these rations. Looks like some example herd that he's got here, its cost on a fresh cow pen is $5.23. Close up dry cows, $2.74 a day based on, on this ration.

It tells us how many feeds you have on hand, so obviously you put feed inventory in there, and so you're looking at what's happening. It gives you some pretty cool cost as well. So feed cost every pen, every day for this, a 2,000 cow dairy, $10,000 in feed cost every single day, $5.56 per cow. So it gives you all of your feed that you have, kind of all of the forages, energy feeds, you name it, you put that all in there, uh, how many tons that you have, and really you can build your own ration from AI.

And it's quite cool, I think, that he has developed this, and I'm gonna have to look a little bit more into this and see what it is. But if you look at the fresh cow pen, it... We wanna formulate it for 83 pounds of milk, 45 pounds of dry matter per day. It tells us it's gonna cost us five point twenty-three cents on a feed cost, and we can put whatever we want, haylage, corn silage, corn grain, molasses, all these premixes, and it gives you so much information on these animals based on the ration that you can use, and it's all been done in AI, which is quite fascinating that you can see all of this information

So there's some cool stuff on his, uh, website, and I'll probably link it there so you can find out what is all there and how to do it. But I stumbled upon this, and I think it's pretty cool, and I'm gonna start exploring it a little bit more. We've been talking about this from a extension standpoint on what we can do to help farmers with AI.

I should say he's got another program on here where you can use USDA's best prediction of lactation records, so it will calculate 305-day and ME milk fat protein and persistency for a herd and all cows, and does it all with AI. You obviously give the AI your, uh, milk production records, and you can feed that through there, and it calculates out what you have for 305-day records Obviously, I've talked to some people about this, and, uh, there's lots of ethical concerns in uploading data and your own dairy data, things like that into AI, and obviously you have to take that into consideration.

Um, I did that with our data, and I found it quite fascinating in some of the things that, that you could find. So that's Padraig's AI tools, and, uh, quite, quite fascinating, quite interesting to, to see that. And, uh, I was quite excited. Been doing lots of stuff with AI. So the next thing I did was, well, looking at our DHI records in our herd.

So we got back all of our test day reports from October, and so it includes herd summaries, heifer calf management summaries, our monthly lactation report, gives back our test day bulk tank reports, udder health reports, everything like that from a DHI summary. So I think there's 10 or 12 PDFs that it gives you, and these are historic records that DHI has a- always put out for farmers, and they've-- they are great records and gives you lots of good information about the cows.

But I would say one of my criticisms has always been, well, I don't know how to summarize all this stuff. You get me 10 PDFs, and I can look through all of them. Um, you may look at them for the first five minutes, and then you just put them in the drawer and, and wait till next month's test. Well, I want to be able to summarize those and see and give me some action items.

So that's what I did. So I put all of these, uh, test reports, uh, and built my own sort of AI summary program, uh, from these DHI reports. I also put our fatty acid profiles in there so I can get a, a measure of what is happening with our fatty acids on milk. So based on our DHI test, basically it tells us we have two hundred and fifty-one milking cows.

We have about 50 pounds of milk, four three fat, three six protein. Gives me a preg rate Gives us fresh cow infections, which at the top I built some things that we should watch. So maybe we should watch these fresh cow infections. We're getting some fresh cow mastitis, maybe watch our somatic cell count, but it gives you some sort of benchmarks that you can look at and where things kind of need attention.

So it gives me some rankings of needing attention. So it says I should look at our chronic high somatic cell count cows and think about what to do them. We should be watching our fresh cow infections. It also says watch our fat to protein inversions. So some had fat percent that was below protein percent.

So maybe we need to look at that for some subclinical rumen acidosis going on, really checking our effective fiber sorting and starch. Now, would I be able to see that on, uh, our DHI records? Absolutely. But here it provides me a good summary. It tells me I have twenty-six of two hundred and forty-four cows, about eleven percent that had fat percent below protein.

Now, it would probably take me a lot longer to look through those records to be able to do that It says our MUN was low, so we maybe need to think about rumen undegradable protein or rumen degradable protein is maybe short in our ration's energy And gives me some things that are right on target. So our somatic cell count was down, so we're making good progress there.

Our new infections in mastitis is kind of on target, six percent, our goal is eight percent or less. And it says that our dry cow period is curing more infections than it creates, and that's a good thing. So our, uh, infections in-- went from twenty-seven percent to thirteen, so that's good. And it also tells us that our de novo and milk fatty acids are above the benchmarks, which means that the rumen is really supporting milk fat synthesis.

So I've made some graphs, uh, so you can graph out all of these things across the different test days to look at our milk production. We've actually been doing quite well, uh, increasing milk production across the last year. Somatic cell count's been going down. Like I said, our new infection rates are... And it gives you some of these, you know, who's the, the culprits in the herd for high somatic cell count and udder health.

So it's giving you those lists, and which those lists are actually on the DHI list, but it's putting them into all one sort of area. It also gives me counts by pens, and then it tells me all my fatty acids. And based on our herds, we can look at graphs. It's graphed them out. Looks like my fatty acids are all on target, uh, for what's going on in our herds.

It just gives a lot of summary information on repro and our herd inventory and what cows are leaving. So repro, mastitis, and other. Talks about our heifers. And then you can give lots of different cow lists here, and so we can see, uh, what's all happening in, in our herd. And so I did this basically simply by uploading our DHI reports into Claude AI, building a nice output, and talking about these things.

Obviously, you've gotta look at those to see if they make sense, and I've looked at all of these results and everything makes sense. Uh, don't see really errors by, by doing this. So I think it's a good way to summarize some of the things that are going on in herds. And I think that in the future, AI is only gonna get crazier, and using it to analyze dairy data, I think, is gonna be at the forefront on some of these farms and more progressive farms.

I'm, uh, very supportive of that. And we're gonna maybe be doing some things, uh, from an extension standpoint and working with farms on how to actually use AI to analyze some of your dairy stuff and what you can actually get out of it. So hopefully I haven't made too many people mad about using AI today in dairy data, but I think it's gonna be the future.

And we're only gonna see this and, um, um, as more AI, uh, gets crazier and using some of our dairy data to do that. So with that, if you have any comments or questions or you just really don't like it, just let me know. And, uh, I'd be interested to hear about what's going on and, and, uh, how we can use AI more to analyze our, our dairy data.

So with that, I think we'll leave it there today, um, and, uh, hopefully you have a good week. Bye.