AI, Satellites, and Smarter Grazing Decisions

Planning grazing by guesswork is costly; planning it with field‑level satellite insight is a game‑changer. We sit down with Chris Knight of Agribot to unpack how AI, radar, and atmosphere forecasting can measure pasture growth, cut labour, and help UK dairy farms make better, faster decisions.

Chris shares his journey from space and defence systems into agriculture, and why the most powerful shift is modelling each field on its own behaviour. Instead of relying on generic equations and perfect weather data, Agribot blends discrete atmospheric states with optical and radar satellites to read how your sward actually grows over time. Cloud cover isn’t a show‑stopper, because radar sees structure through the gloom. The result is consistent, frequent insight: near‑term dry matter estimates and multi‑week scenarios that flag above‑ or below‑average growth so you can adjust rotations, fertiliser plans, and buffer feeding before the pinch hits.

We also get honest about “accuracy.” Plate meters are useful, but they’re not a gold standard. The win here is consistency and context: a system you can calibrate to your farm that shows change early and keeps watching every field. That opens a bigger conversation about metrics. Should we keep chasing kilos of DM per hectare, or shift towards usable grazing days, megajoules per hectare, or even expected milk from pasture? Different systems need different lenses, and the same data can support them all. What matters is turning complex signals into decisions that raise milk from forage, protect margins, and reduce stress.

There’s a wider story too. With labour tight and weeks long, measurement is the job that slips. Tools that save half a day a week without sacrificing control can lift farmer welfare as much as farm performance. Chris’s team has kept the work mission‑led with support from Innovate UK and Horizon Europe, proving models before monetising and partnering with early adopters to refine accuracy. If you’ve wondered whether satellites can truly help you manage grass in a cloudy country, this conversation gives you the how, the why, and the next steps.

Enjoyed this? Follow and subscribe, share it with a mate who still plate‑meters every paddock, and leave a quick review so more dairy folks can find the show.

For more information about our podcast visit www.chewinthecud.com/podcast or follow us on Instagram @chewinthecudpodcast. ChewintheCud Ltd is also on Facebook & LinkedIn. You can email us directly at podcast@chewinthecud.com

Chapters

• 0:00 Welcome And Autumn Grazing Review
• 3:20 Why Grass Planning Starts Now
• 5:18 Introducing Agribot’s Grass From The Sky
• 5:28 Labour Saving And Consistency Payoff
• 5:48 Guest Intro: Chris Knight’s Journey
• 5:58 Sponsor Messages And Show Intro
• 9:40 Choosing Agriculture And Its Impact
• 10:00 Atmosphere vs Weather: The PhD Insight
• 13:50 Discrete Atmospheric States And AI
• 16:15 Field‑Level Models Beat Generic Models
• 18:00 How The Satellites Actually Measure Growth
• 21:10 Mixed Swards, Soil Types, And Learning
• 24:05 Farmer Welfare And Time Saved
• 27:20 Why Agribot’s Ensemble AI Is Different
• 29:36 Forecasting Growth Two To Four Weeks Out
• 33:05 Real‑World Use: Planning For Wet Or Dry
• 35:05 Adoption, Insurance, And Market Interest
• 38:40 Grants, Horizon Europe, And Being Mission‑Led
• 43:30 Accuracy, Ground Truth, And Plate Meters
• 48:10 Towards Better Metrics Than DM/Ha
• 52:00 Quality vs Quantity: Energy And Milk
• 56:20 Consistency Over Perfection
• 1:01:00 Closing Thoughts And Takeaways
• 1:10:40 Legal Disclaimer And Credits

Transcript

Andrew Jones: 0:14

This is the Chewing the Cud Podcast, a podcast for the UK dairy industry, brought to you from the southwest of England and listened to around the world. Hello and welcome to Chewing the Cut Podcast. My name's Andrew Jones, and with me as usual is Sarah Bolt. How are you doing, Sarah?

Sarah Bolt: 0:40

I'm very well, thank you, Andrew. And how are you?

Andrew Jones: 0:42

Yeah, not too bad, thank you. Not too bad. So today we've got quite an interesting subject, I think.

Sarah Bolt: 0:49

I think it's a very interesting one.

Andrew Jones: 0:51

I mean, i cows are only now coming in as we head into winter, and I suppose you could say this is for next spring, but it's gonna make you start thinking, I would have thought. You to be honest, you should be thinking about your grass for next spring now as you take the cows off, shouldn't you?

Sarah Bolt: 1:05

I was gonna say we've had some good autumn uh grass growth, definitely, haven't we?

Andrew Jones: 1:09

Uh yeah, a lot of clients I've spoken to, you know, they've got the cows in now because it got wet. Um, I know it's been a little bit drier this last week again. They got them in and they're saying we've got loads of grass out there still. Probably going to affect intakes as it was getting wetter, so you might um not quite be getting energy in the cows, but they're all saying how well the cows were milking, how the well the solids were doing it just off the grass. So we've had very good autumn grass, definitely. But as we finish the grazing season, we should be setting it up ready for next season.

Sarah Bolt: 1:39

Yeah, getting that wedge uh all ready for next year.

Andrew Jones: 1:42

Well, exactly. I think they could be uh that there could be happy days for some sheep this year with uh the amount of grass that's around.

Sarah Bolt: 1:48

But anyway, let's we talk only if we can get them off by Christmas.

Andrew Jones: 1:51

Well, exactly. The less we talk about sheep the better. Um as I would say, they're good for eating, but there you go. Um but anyway, sorry, make me think about college, college friends who had sheep always used to describe them as woolly lemmings, but that's another story. Um anyway, so today, yes, today I came across this. Uh I saw a social media post actually, just uh an invitation to the uh Southwest Dereo Development Centre, uh, and our speaker was there giving a presentation on the work that they'd done. Um and well, as we were discussing before we came uh to record this, wasn't it? It's it's potential um there's lots of advantages, but potential labour saving that this could bring and save someone potentially half a day a week.

Sarah Bolt: 2:33

I think that's the the key to this, really, isn't it? It's the the accuracy is there and uh the the half a day a week labour saving. I think most farmers would uh grab your hand off at that one.

Andrew Jones: 2:45

Well, exactly. And it's it's the consistency, isn't it? I mean, uh I guess it's the point we discussed, is how accurate it is in the podcast, but it's the consistency of it, and and ultimately it that's what counts. And as you say, if you've got a reasonable farm and you've got to go and walk it every um uh weekly, by the time you plate meter it, even if you don't plate meter, you just walk it, there's probably half a day a week gone, and you'd make some management decisions in what you're doing. So hopefully this will bring some consistency and allow some labor saving um to concentrate on making those management decisions.

Sarah Bolt: 3:16

Definitely.

Andrew Jones: 3:18

Good. So, anyway, let's go find out more about uh Agribot and its uh measuring grass from the sky. This podcast has been brought to you today by TuneTheCud Limited, who offer completely independent dairy and beef nutrition, our signals advice and training along with ROM's mobility scoring. For more details on these and other services available, please visit our website www.tuneTheCud.com or email us directly on nutrition at tune the cud.com. TuneThe Cud Limited now offers first aid training from a registered first aid at work trainer and experienced minor injuries practitioner. For more details, please visit our website www.tunethecut.com or email us directly on training at tunethecud.com. Hello, I'm Andrew Jones.

Sarah Bolt: 4:20

And I'm Sarah Bolt.

Andrew Jones: 4:21

And welcome to the Tune the Cud Podcast, a podcast for the UK dairy industry.

Sarah Bolt: 4:26

Farmer, advisor, processor, and everyone else. We have topics and episodes that will interest you.

Andrew Jones: 4:31

We discuss the practical and the technical aspects of different UK dairy industry topics.

Sarah Bolt: 4:37

We aim to make you think about what you're doing and ask yourself, can it be done differently?

Andrew Jones: 4:41

Listen to us speak with specialists from inside and outside the industry about their area of expertise.

Sarah Bolt: 4:48

Subscribe and listen to episodes for free on your favourite podcast platform.

Andrew Jones: 4:52

Sign up to our website, www.tunethecud.com for podcast notifications so you never miss an episode.

Sarah Bolt: 4:59

And links to our socials, including Instagram, TuneTheCud Podcast. All one word and remember, no ginny.

Andrew Jones: 5:06

Or email us direct on podcast at tune the cud.com.

Sarah Bolt: 5:10

If you like what we do, please share and leave us a review to show your support. And that's it.

Andrew Jones: 5:15

Enjoy today's episode. Hello and welcome back to Tune the Cut Podcast. Our guest today is Chris Knight from Agribot. How are you doing, Chris? Oh, great. How are you? Oh, not too bad. Thank you. Not too bad. Now, today we're here to talk about grass measuring with the Agribot system. But uh Chris, just before we start, tell us a little bit about yourself and how you got to where you are today.

Chris Knight: 5:48

Uh well, I'm a software developer, if nothing else. Um I started off um at eight years old, doing various things for my teenagers, but then I studied AI and robotics way back before it was trendy, so back in 2005, actually. Um after that, I worked in space and defense, working on various satellite constellations for various governments. After that, I worked in uh as a software developer in high frequency trading, insurance, and all those kind of things. I was kind of a typical software developer. I didn't care which sector I worked in. I was more focused about working with particular technologies and got a job, whatever was doing that. Then about five years ago, I decided I was going to flip that and chose a sector to work on and trying my best in that sector. And I chose agriculture. Um so then at the same time I had some ideas around how to use AI for doing long-range forecasting and agriculture. And I got uh I formed Agrobop then. Um, but then I also got roughly around the same time I got approached. Uh well, I I was in conversation with Cranfield University about some of ideas, uh, and they uh talked me into, although it wasn't very hard, to turn parts of it into a PhD. Um I wasn't looking to do a PhD, and uh but I ended up doing it, uh, for which I finished that uh two, three months ago, actually. Um I'm not sure if I'm officially doctor yet, but hopefully soon. The certificate's not gone for yet, so I don't know where that sits. Um and that was around doing uh using AI and atmosphere to do long-range forecasting, which then comes into Agribot Agrobot. So, of course, Agrobot's just a columnation of all these things AI, earth observation, long-range uh modeling, put them all together and off the company rent.

Andrew Jones: 7:43

Off the company rent. Now, now I met you back earlier in uh in the year when you were giving a presentation, weren't you, at the Southwest Dairy Development Centre, um, which I'll be blunt and say I was a bit disappointed, the lack of farmers there, because I think it's some great stuff when you actually hear about it and uh hence why we're here. Because it's what you've done is a bit different. I mean, I've I've I think I might have said on the day, I've sat there for five, possibly even ten years. I can't remember how long it's been there, going, when are we going to get grass measuring via satellite? Because Southern Hemisphere's had it for a fair old while, and of course, they rely on um chloroform density. My understanding, that's how they measure, they measure the chloroform density, and so the more chloroform there is, therefore, the more grass there is, and that's how you measure grass. The advantage I guess they've got compared to here, as I look out the window, is um, yeah, we've got a little bit more cloud cover than they usually see, which I believe is one of the problems why it hasn't happened here. But I suppose take a step back from what you're going. Why choose agriculture? From somebody that doesn't have any background with agriculture, why go, I'm going to work with agriculture. You know, most people sort of look at it as, well, hard work or or or whatever it is. What made you choose agriculture?

Chris Knight: 8:54

Well, I'm not from agriculture, but I grew up in uh, you know, uh not quite rural, but rural enough part of Scotland, south of Edinburgh. So I some I grew up with some farmer, so I wasn't completely disconnected from it. But yeah, I personally don't. And why agriculture? There's a culmination of things. Um I I think because I was saying thinking, I was the first time I actually sat and thought about what I want to do. And of course, everyone wants their life to be impactful, uh, whatever, however, you define that. Um, and I suppose my definition sort of went along the lines of what can be more primary than uh helping the food supply system. Like um, it's can't get much more important than that, really. There are there, of course, there's other important things, but that's that's up there with that's up there with most of them. At least exactly.

Andrew Jones: 9:44

They say the um the start of any advag advanced civilization is agriculture, isn't it? If you don't have that, then people can't go and do other things. So so but but your PhD, you probably ought to explain just a little bit more about what your PhD was because that's quite relevant to what we're talking about, isn't it?

Chris Knight: 9:58

Yeah, so the PhD was uh it it that came from that initial little fact that um a lot of yield models, agriculture lots of agricultural models, uh require weather as an input weather. That's typically sunshine, rainfall, humidity, stuff like that. Uh and as we all probably more aware than we'd like to be, uh, if you're planning on barbecue, you're being quite brave if you listen to the three or four-day weather forecast. So, of course, that's the limiting factor for all these uh models. Um but as a science or as forecasting, not that I forecast weather or atmosphere, but we're actually better at forecasting atmosphere longer range, like in the several weeks. So that's when you're watching the longer form weather uh reports when you start talking about highs and low pressures. That means sea level pressures were actually really good at forecasting those further out. We're just not very good at turning that into weather. So then that that bore the question, well, why don't we use these as inputs for agriculture as opposed to using weather and make advantage of that?

Andrew Jones: 11:03

Well, I was gonna say maybe I'm showing my age, but it feels like the weather forecasting is getting worse, not better. The fact that they've got you know all these supercomputers, they seem to get it more uh wrong than right these days, but maybe that's just me. Maybe.

Chris Knight: 11:19

I I don't know what to say to that. I'm very insulting. I I can only insult someone by answering that properly.

Andrew Jones: 11:28

Well, I mean, uh you know, you might be too young, but you know, we all remember uh Michael Fish. No, there will be no hurricane tomorrow.

Chris Knight: 11:35

Yeah. I've actually I've actually used that as a I've used that as a a problem with AI in a sense. So I do some lectures for on on AI as well, as my background, if anything. And I often use that as a good example because it wasn't so much that he was wrong, it was the confidence level. So AI often gives a prediction, but good models also come along with, well, I think it's this, and I'm about 90% sure. Uh whereas I think Michael Fish came out and said, This is happening, there's no chance of anything else happening, and then that was the issue. So I was I was saying it wasn't a prediction, it was the confidence level that was the problem of that one. But um, you know, but I think we've got better. I would the science would say we've got better at it, but of course I'm not gonna I'm not gonna argue with anyone, I'm not gonna argue with anyone's lived experience as well, because that's a very they like they they like to think they're more accurate, but I'm not gonna comment on the real world.

Andrew Jones: 12:30

I wouldn't say I'm accurate at all, it's just that's the way it feels, but then I don't bother watching the weather these days, it is what it is. Um but um um so yeah, so your your your sister, your your postdoctorate, sorry, I should say, is using atmospheric pressures rather than weathers. Is that the right way to describe it to then predict what's happening?

Chris Knight: 12:52

Yeah, I mean I'll get jumped on for saying this, but to make it if you're predicting the weather, you'll have to look at the atmosphere. So it's more complicated now, of course, is before anyone jumps in too hard, but it's for the sake of keeping it simple. So they'll look at the weather people will look at the meteorologists, not the weather, but the murologists will look at the weather, weather people as well as but they'll they'll look at the atmosphere and then they'll do their forecasts based on that. Then, of course, their forecasts then uh go into agricultural models, and then so you're you're you're doing multiple layers of models, every model has its inaccuracies, so the more layers you have, it compounds etc. And the good the the good phrase for AI fields is every models every model's wrong, but some are useful. So the more models you have in there, the more so with doing atmosphere, we sort of make like if if to pick on uh I say to pick on it, but it's actually good for this podcast. If we say grass, so rather than having grass as the third or fourth model down, we actually put grass beside weather. So we don't forecast weather anywhere, we're going from atmosphere to grass rather than atmosphere weather grass.

Andrew Jones: 13:58

If I remember right from your presentation, did you say something like only 28 or 29 different atmospheric models that you can then use to predict? Yeah, right.

Chris Knight: 14:07

Well, so the UK MetOffice they do a their long-range weather forecasting, it's called uh fluil decider one. They're they're based on 30 states. So the idea is about making the atmosphere discrete measurements because AI loves discrete measurements. AI doesn't like like temperature, for instance, is a continuous value, right? It can be anything between any values. AI, again in general, doesn't do so well in those scenarios, but if you can just say here's categories, it does really well in that. So it's about turning the atmosphere into categories, and to do that, we say any day belongs to one of these 30 states, and those 30 states represent every state the atmosphere has been in the UK since like 1860 or something like that. So every day since 1860 belongs to one of these 30. And of course, then that bees into the AI, because then the AI likes that kind of categorization and off that goes.

Andrew Jones: 14:58

Yeah, that it knows that state, I don't know, let's say seven is going to predict it based on previous experience, will produce this much grass growth or whatever it happens to be, uh ultimately is what's coming from it.

Chris Knight: 15:10

Right. So then uh to go to a little bit deeper how agribut works, so we use Earth observation. Sometimes farmers have their own measurements, but we can go into maybe measurements later because there's there's a lot to go around how about measurement problem. But let's just at this point say the measurements are accurate. So then we can get the farmers' measurements, then we know the atmospheric state all the way back. I suspect our work records for atmosphere is longer than the farmers given it's 1860, but yeah, you know, there might be a farmer out there somewhere, I don't know. Um, so let's say to go back to 2015-2014 is quite common one. We can actually get the satellite data back to 2016. So then we can see how every how a particular field has behaved under each of the different atmospheric states. Then we can build a model down to the field level. So the AI models actually learn how an individual field behaves rather than using these more general models, which you get a lot of vanish. When you use these general models, that's when you have to give it the soil type, the particular grass type and whatnot. We don't typically need that because it's it's the model's learning how the field reacts to atmosphere, not a general model.

Andrew Jones: 16:15

Well, say you say about that. It's I know we're sort of jumping a bit, but it's quite amazing. I remember your presentation, you basically have got the whole of the UK mapped, haven't you? So you know what's growing in each field. Just was it off the growth curve of the plant or something? Was it?

Chris Knight: 16:29

Yeah, so it depends on the well, it doesn't depend. So REML, you can imagine we have a pipeline. So we we identify all one point depends, but 1.6 to 1.8 million fields in UK. And we measure the data coming in um as it comes in. Uh we do that because we've rewritten a lot of the software that does that processing. So one that we do it a lot faster because we can. Um rather taking 18 days, it takes us an hour and a half to do it. So that's why we can do it per field. Um and yes, so we we track every individual field. Again, we we focused a lot on grass recently because there's a lot of focus on grass because it's for a long time, I think you alluded to it, it was often the forgotten crop for quite a while. So there's lots of opportunities to then address that. So that's partly why our our uh we've had a lot of attention on grass.

Sarah Bolt: 17:20

But so I'm gonna go back to basics because Andrew knows what Agribot is. I know a little bit less as to what Agribot is. And you're saying that you're collecting this data from the fields. How are you collecting this data? What's what it what is it that's collecting the data?

Chris Knight: 17:37

So from a data perspective, you can see it, you can split split up into three sections. So we do real historical real time, we sit next to each other. So that's how it's behaved in the past and how it's behaving today. Uh, that's typically more useful for farmers uh because there's a data gap sometimes between the field and the farmer just purely because they don't have enough time in a day to visit every field every day in our main once a week. So that's that part. So that's the part we use uh satellites for. And uh it was mentioned earlier that the core films of the field they infer it. So we use optical, which is interesting, it looks at the near infrared of the crops um and the green. So it doesn't look for the colour films particularly, but it can tell by the difference between the near infrared and red, what how they're behaving. We see that when crops go from green to yellow, that's a drop. So we see that, but that change happens in the near infrared first, which we don't see, but the satellites do see. Um but the issue I was alluded to there, if it's a cloudy day, you don't get measurements, and which is a massive issue with us. But we also use radar um satellites, which they're more looking at uh how much physical things so the first ones are looking at colors and they're looking straight down. Radar satellites are more firing a beam down and then measuring how much comes back. So you can imagine if it's a small bush, a small signal comes back to the satellite. If it's a bigger bush, a lot of it more comes back. So that's sort of looking more at the physical attributes. So to do the land classifications, every crop grows at different speeds at different phases. So we can identify the crops based on we'll we'll we'll fire the radar down and we'll over a period of time we'll see a curve of how that's changed. And that curve tells us how it's growing, and how it's growing tells us it's kind of crop. And then, of course, once we know the kind of crop, we know what how it should be growing, and we use another AI model to make sure it's following close enough to that graph. And if it doesn't, then we know that there's a health issue uh with it. Um, or if we use it, if we do have the optical with the core foams, we can actually see that there. But we don't we try not to rely on that because there's a lot of clouds. That's this I think it's like 79% of days the satellites go over. There's clouds. So yeah, it's a bonus to get it, but not it's not so what about sort of mixed crops?

Sarah Bolt: 19:54

If you're thinking of same, if we're talking, we're talking grass, so multi-species lays, does that have a an impact on what you're looking at?

Chris Knight: 20:02

And if we're doing if we're doing growth, not so much, because again, it's relative to the field. The only time you can get tripped up a little bit is if you change if you if you're historically growing a particular grass, then you suddenly change to it, um, then it can it becomes less accurate for a little while while it learns that the change has happened. Yeah. Um, but what the longer you keep it that, that'll it'll correct itself after well. And I say less accurate because it's still accurate enough to be useful uh to be more than useful. But um, so it does impact it a little bit, but not much. And I but again the reason why the these why uh the type of crop often matters in traditional models is because the inputs for the model are the weather, temperature, rainfall, whatever, then you have to tell the the type of uh grass it is, because of course different types of grass respond differently to those inputs, and then you get your output. But because the models feel specific, that's why we don't need to worry about the type of crop. In fact, sometimes we can actually infer what the type of crop is, because we can see again, we look at the curves, that will sometimes if the difference is big enough between species, we can actually tell the the subtype of crop that's in there.

Andrew Jones: 21:14

Yeah, because it was it was one of the things that amazed me when you did the presentation. I said, Well, don't you need soil type? And you went, Well, no. It was like, oh, okay, you must, yeah, because I was thinking you must link this with the you know, the soil maps for the whole of the UK, aren't there, in terms of what's where, etc. And you're like, no, don't need it because we're we're not measuring the data that, like you said, traditionally models use. We are looking at that that growth phase and using the weather to protect.

Chris Knight: 21:41

Exactly. It's the soil types of the other input, right? So you've got the weather inputs, then you've got the crop type, then you've got the soil type, and those are your typical inputs to your traditional models. But of course, because we have a model per field, yeah, it it compares itself to the field itself. So we have the model specific to that field rather than yeah, and that's that's the point. Um which is which is a massive game because we're fundamentally, you know, there's also ways of viewing what we're trying to do, but it's a quality of life tool. If you can view this as a quality of life tool in the sense that it's trying to save you time, so you don't you don't save someone time by asking them, you know, what's their shoe size and all those other questions to do the modelling. That it may end up costing them more time. I I think when we're talking, I was I've just come back from a trip actually, again, talking about um all these, what's agri tech's future and all this? Uh uh, you know, I think it leans into what I've just said is that there's lots of opportunities, and we're being agri tech sectors being focused on like animal welfare, environmental welfare, all those kind of things, which of course we should be. But I think one thing that's often forgotten is the farmers' welfare as well. So um, everything's forgotten about, but you know, when we talk about feedback for the people we're helping, you know, it's often they'll talk about how it's helped the farm's profitability, how it's helped the cows or it's helped this or that the fertilizer use. But the biggest one for me is when you hear them talking about what's allowed us to go from a six-day week down to a five-day week and all the benefits that gives to the individual. So, you know, and I think that conversation went well over the last few days. So I think in the future these opportunities are going to start focusing more on uh farmer welfare rather than just the farm's welfare.

Sarah Bolt: 23:20

I think labour saving is a huge one within the industry here in the UK, isn't it? That labour is in short supply and therefore anything that can save labour is definitely worth looking at.

Chris Knight: 23:32

Yeah, and that's where they're connected, right? Because if there is a there's a completely separate farmer discussion uh in the sense that just the farmers themselves are having to, there's just like we said there's just not enough hours in the week or the month to do what they want. And that brings a certain amount of stress, but then that's amplified by the labor shortage, which is a separate but connected issue because then that magnifies the first problem and it begs builds more stress and it builds more time. And you know, a lot you ask most people around the world to take a six-day working week, or some even do a seven-day working week. And uh, I think a lot we're we're starting to try and focus on uh well, we've always focused on it, but trying to find uh other ways of incentivizing other agri tech sectors to also have farmer welfare along beside the animal welfare and all those kind of things, it should be a zero-sum game. We should talk about welfare across the field. Um, so that's one of the big impacts we like to talk about.

Andrew Jones: 24:26

I mean, you're talking about time, let's be honest. If you've got a you know, reasonable farm to um walk all over, it's probably half a day a week, isn't it? So by the time you plate meter the whole farm, input that data, make some management decisions based on that data, you might as well wipe out half a day for the week if you want to do it. So, as you say, if you're saving somebody that half a day to and it's constantly being monitored, so you you know you're you're not having to stress, oh, I've got to do that this week, or it, you know, it's one of those jobs that might slip because of other things, it makes a huge impact, doesn't it? That they can use that time more productively. Right.

Chris Knight: 24:59

I mean, we'll often talk about it's one of the critical jobs that gets not that isn't done. Because of course, when you're a farmer, um it's not about the jobs you should do, it's about what how many critical jobs can I do? Never mind the ones I should do. And you know, I think measurement's one that often gets uh put by the wayside. And it in a way it's a tricky one because it's saving time on that one's a false economy because the more you utilize your grass fields, your pasture fields, the cheaper it is to produce, say, milk. You know, if they're out feeding grass, you know, per your price per litre goes drastically down and compared to in the shed. But you can't do that properly unless you're at measuring. Then of course, you know, again, it it it it's uh it's a no-win scenario for farmers in terms of where to put their time, right? So we can talk about that and they should be doing it. But then there's a bunch of other stuff they should also be doing.

Sarah Bolt: 25:48

So Yeah, I was gonna say, Andrew and I have probably spent many a year telling farmers that uh they should be going out and and using that half a day to to adequately uh measure their grass and and make the most from it.

Andrew Jones: 26:01

It's uh I used to joke, someone used to pay me to go and uh walk their fields for them once a week, uh local to me, which I quite happily did, but I was like, well, someone pays me to do my weekly exercise, so I'm not going to play.

Chris Knight: 26:14

I mean, some people do love it as well. And if you if they if they use our tool, they can't I would encourage them to continue doing it because then that completes the food the the cycle back in the model gets more and more accurate. But of course, we're also there to not make them do it as well. So we're not yeah, so for some farmers it's great because they don't have to, but we're not trying to stop the ones that do want to do it as well.

Andrew Jones: 26:34

So so so taking a step back a little bit, you've sort of explained that you know what's in there to get it, but what makes yours different than what's gone before? Because I know since I've met you, there's a few times I've gone, oh, I've met this guy and he's got this system to measuring grass, blah blah blah. Oh well, we've tried that before and it doesn't work. So so what's different about your system? I guess one, the fact that you're not worrying about cloud cover, but what makes your system different and more accurate than maybe some that have been trialed before?

Chris Knight: 27:03

It it's the one the thing I often miss under talk uh because it's more of it's often comes across as a computer science job, but it really is that model per field basis. That's where a lot of the accuracy comes from. And I think the other, you know, just to talk about AI a little bit, and okay, the way that people build AI is they want to build this monolithic AI that's amazing at everything. Um but the reality is that's not a great way to go forward. So instead, what we do is we we do what's called an ensemble approach. So as much as I've downtoped the traditional mechanistic models, that that's the ones that you know, the traditional ones that we've all used. Rather replacing it, our models are built to build alongside it. So rather than one model trying to do the forecast and the prediction, we have multiple models that all have their own little specialty. So it's almost a committee of models. And then we have a model that sits on top of that to decide which one it's going to listen to. Uh, it does that because that sits and looks at the weather that's happened recently. It looks at and it figures out which models tend to do better in those conditions, then it slightly listens more to that one. So um you can it's tricky because I don't want to answer for people's if these are companies, uh often it's companies that I don't really know how to do it because of course they don't tell me. Uh I I ask, but fair enough, they don't, you know. So I don't want to say, but I know how the academic and all the published models work. And I know how AI people tend to think and they want this one big model, but we took a leaf from uh weather's for weather forecasting does the exact same thing. So they have different models that are slightly tweaked to be better at different things, and you sort of take the average or the the popular opinion, and that's kind of what we do as well.

Andrew Jones: 28:50

I guess you could say it should be that's what should be on farm almost really. You rely on the advice of your vet, your nutritionist, your um uh foot trimmer, whatever it is, to then try and build up a bigger picture, isn't it? And and and then get the answer. And effectively that's what you're saying, I suppose. You can say each each model is specialist in their thing, and you're just bringing it all together to to uh to make that prediction. Exactly.

Chris Knight: 29:13

Um and we get a lot of accuracy that plus you know, we're we're as far as I'm aware, they're the only ones that really do atmosphere. So when you start to get to the two to three, when you get into multi-week one, we have this advantage of uh, you know, my academic research feeding into that. Um so there's there's that aspect as well.

Andrew Jones: 29:32

So how far in advance can you predict grass growth?

Chris Knight: 29:36

Uh I I typically two to three weeks I would say is we're pretty accurate. Four weeks is good. Um you know, once you're at four weeks, but once you do multi week, it I think the other thing that it is caveat is that we only typically give you a number of giant matter mass up to four or five days. After five days, it tends Be more categorized. It tends to be like this it's going to be a good day or a bad day. And then we give you a rough estimate of what that typically looks like for that field. But after a week it turns more probabilistic. It says it gives you pers it gives you scenarios, it gives you certain odds. So it says the odds of it being in this range or the odds of it being that range. Which, you know, if you're doing three, four weeks, that's kind of normally what you're looking for anyway, right? It's through three or four weeks tends to be a blind guess, but at least this is a this is a bit of data to back some sort of decision.

Andrew Jones: 30:33

And if you know you're going into a dry period or then that's when you can start slowing the rotation down, start bringing some buffer feed in or whatever it happens to be.

Chris Knight: 30:41

Right. So I mean accuracy, you can discuss that at length in the AI field. So accuracy tends to be like the percentage you are away from the actual value. Um but in this instance, if we're talking three, four weeks out, it's above 90%. But the 90% is saying, is it going to be amazing, good, average, below average, really bad? And the accuracy of it is when it's wrong, it's normally one category out, uh, as opposed to being completely wrong. And one category out is typically going to be like it's wet versus really wet. It I can't remember ever thinking it's going to be wet and it turns out to be dry.

Sarah Bolt: 31:17

And I think just that amount of accuracy over that length of time, I mean I think any farmer would probably grab your hand off of that, knowing you know, the approximate weather conditions for the next three weeks will just allow that planning process.

Chris Knight: 31:34

That and that's the that's the vanish. So I again the the downside of if you look if you take a step back from agriculture, the downside of doing um atmospheric forecasting is exactly atmospheric forecasting isn't about like if we took two two to three day forecast, that's normally about saying it's going to rain at three o'clock on this uh on Tuesday, it's going to start raining at one o'clock. So it's trying to be very precise, which is for our day-to-day life, that's what we want. But if you imagine two to three weeks for a plant, you know, I just want a certain amount of rain over a certain amount of period of time. I'm not interested if it's two o'clock or four o'clock. So only as I get this amount of rain over this period of time. And that's what atmospheric forecasting is really good at. So that's why it lends itself really well to agriculture. Atmospheric forecasting is rubbish for helping you plan your barbecue. You can't do that. So don't email me. Don't email me about asking about barbecue in four weeks. I'll get I can give you the odds on it happening raining over two days. But um, yeah.

Sarah Bolt: 32:29

But it's that overall growing period, isn't it? It's it's is it worth, you know, have I got to to shut off some fields? Have I what have I got to do over those those three to four weeks?

Chris Knight: 32:41

And well, exactly. And I think that's why it's perfect for agriculture. Like I bring this up because uh part of the pushback we get is it sounds like we're saying we can do three to four week weather forecasting. That's not what I'm saying. Uh we're doing growth forecasting three to four weeks out, and it's probistic scenarios giving you the odds to place your bet on the fertilizer game, I suppose.

Andrew Jones: 33:05

But that makes perfect sense, doesn't it? The closer you get to the the the day you're looking for, the more accurate it's going to get and the further away, as you say, it's the probability is it will be wet or very wet, but it might be one rather than the other. But we're just that's where we think it is. But ultimately you can look and go, well, it's gonna be wet in a month's time, right? What do I need to do? Or it's gonna be dry in a month's time, right? What do I need to do? And then make your decisions based on that. It helps you make those decisions, doesn't it, ahead of the game.

Chris Knight: 33:34

I mean, there's not a lot of difference in preparation between wet and really wet, right? I mean, but but once you get within a week, that's when you start to know that that's when we start honing in or actually which one it's gonna be. But you know, the big difference is well, if you prepare for wet, if your bet isn't wet and it turns out to be dry, that's when you get caught out and that's when you're in trouble. Um, but that this stops that.

Andrew Jones: 33:55

Um so what's so as to say, you mentioned pushback, what's the reception been? Obviously, I met you at the the talk at the Southwest Area Development Center, but I'm guessing you've been elsewhere and and uh given the presentation on it. So uh what what how how do you find it?

Chris Knight: 34:08

Yeah, really good. Uh your listeners can't see it, but I do, I'm sure pointing to the bags under my eyes. That's a that's a sign of how well it's doing. Uh it's doing really well to the point that uh we're we're trying to uh keep up. I mean, we we already have uh amazing farmers who signed up who sadly I haven't reached out to for a little while, but that's because we're busy trying to get the models ready for the next going season. So if they're listening to this, we're still turning away, we're gonna be ready for you. Um just but the but it's all up and down the field. Like for instance, the other people who are interested in this is things like parametric insurance, uh, which is a much better type of insurance for farmers, because there's if something goes wrong, you're not stuck in a lengthy conversation with insurance companies. They don't like the conversation either, to be quite honest with you. So, parametric, once we understand how the atmosphere affects your field, we can then say, well, the probability of that happening this year is this. Uh, and if it does happen, that's an objective uh trigger for insurance payout right there. There's no discussion around it. Uh, there's lots of interest there. So there's interest up and down the field, um, so to speak.

Andrew Jones: 35:16

Uh well I guess you you you made the comment I remember at the thing as you said, I've come to help dairy, but he said I could you said I believe you said something like I believe I can make money, more money out of other things, but I want to help dairy at this point in time and the other stuff will come.

Chris Knight: 35:30

Exactly. I I think the uh if we're talking if I'm talking to potential investors and whatnot, I always make the point of you know, I want the company to be a roaring success, but I don't think that's going to be in the back of farmers for two reasons. One, we don't want to, we don't believe that's the right way to go forward. And two, I'm not sure there's that farm that much money to be extracted from it anyway. So we're being both kind and all and realistic on it. So that said, with the farmers who joined, and we're we are still taking on board, the farmers who come and help us and provide our data to help us improve their model in the early days, they get to use the product for free and perpetuity right now. Because as a thank you, because it, you know, they're also helping us. If the the theory being that if we can keep farmers happy with our predictions, then no one else can really moan. Uh, and the and the wonderful thing about farmers as users is they're very quick to tell you when it's when it's not good, which is what we want. Uh, you know, it's it's wonderful hearing how good we are, but there's more value in and more value in hearing when we get it wrong.

Andrew Jones: 36:30

So, how many, how many farms have you got working with you at the moment to what's the word proof the data?

Chris Knight: 36:35

Uh 15 or 20 farms. Um, but that's great, that that's hindered again by me just uh by the team trying to onboard them as quick as we can rather trying to build that out as well. But from an accuracy perspective, like you know, we've we're happy with it. Like the other reason why we've not taken money from is like how I keep joking that we're an old-fashioned company and that we'll we've we spend time building it and proving it before we take people's money, which uh, you know, and the modern way of running tech startups is counter to that. Um so we're happy with accuracy now, to be quite honest. And it's gone through academic rigor, all those kind of things. There's papers published on it, so we have spent more time than uh a lot of other tech companies uh proving out. And now it's just about building the business uh now that we feel comfortable about uh you know what we're seeing and uh you know making a product out of it.

Andrew Jones: 37:28

So when will you commercially come to the market with the product?

Chris Knight: 37:31

We're really trying to get it up and running for the start of uh next season, which you know, February, March.

Andrew Jones: 37:38

Yeah, so it it is really all systems guys.

Chris Knight: 37:40

Yeah, we're uh we're all goal. Like we're we've actually so we've spent all our time validating and creating the models. We're actually working on uh the actual bit that automates it. So right now we we sit and look at every individual field and we go through the system bit by bit manually, and now we're comfortable, we're happy with that, we're replacing ourselves so it's a lot more automated. Uh so again, keep we keep the thought of saving farmers time in our heads. So we're putting a lot of work work and effort into minimizing that. Again, I don't want to. It's one thing being accurate, but we've got to be useful as well. And we get we're useful if we save people time.

Andrew Jones: 38:20

Uh like we, I guess, seem to have that conversation with a lot of the innovators we get ag tech innovators we've had on the podcast, because we say we made a comment recently. It seemed to be a lot in the UK at the moment, which is grand. Um, any interest from overseas?

Chris Knight: 38:34

Yeah, there is. Um, lots. Like we just we were leading uh EU Horizon grant, which um you know just takes that's like the e when the EU talks about how many billions it's spent on innovation, it goes through Horizon. So we built a consortium of 13 companies to do that. We also just won uh an EU um EIT grant, so EIT food. It's a subdivision of the EU. Um we won an innovation loan from them.

Andrew Jones: 39:01

So they're still talking to you then. As in we're, you know, we're at we've we've done Brexit. I thought we, you know, that wasn't allowed anymore. Well this is not as a podcast in the store. Wow, yes, right now. I'm not trying to stir the pot there as such, it's more tongue in cheek.

Chris Knight: 39:18

Uh no, the the the short answer is we are still part of well, can't be careful of that. I think we're we're part of Horizon, which is part of the EU, but then so's New Zealand, and so's Canada. Okay. Um, so Horizon was is originally an EU thing, and it is an EU thing, and it's funded to the benefit of the EU. But certain third countries can join. Of course, the UK was one. We were out for like a year, but then we came back in, and then um New Zealand joined because of course they're big dairy uh agriculture place. Uh so of course they should be part, but the horizon is more about agriculture, as I bring that in. But uh they do have an uh agriculture focus as well. So yeah, they do like us to a degree, like they don't they don't they don't like seeing if so I talked about consortium. If we have a consortium with 13 UK businesses, yeah, they're not gonna like that. Um, but then it is designed to be collaboration across the whole of the UK. Uh the EU. But you you mentioned uh why we see a lot of agri tech, because we actually we're conversations with all American companies, and to be fair to do there's a lot of hate uh going around for the state of the UK and the EU right now, I suppose. But I think you touched on it, just be positive about part of it, and part of the reason we're seeing a lot of agri tech companies is there is a lot of uh grant money out there for it. We're the beneficiaries of it as well, and we we you see that no better than when we're doing our international collaborations. So we would we meet an EU or a UK business, we meet, we like each other, we think great, we can collaborate with you. All right, well, what grant can we go for to help pay for this collaboration to make a better product? And that's the conversations. We're also in talks with American companies where it's a lot more difficult. So we we we really like each other, we really want to uh collaborate with each other, but we're small businesses, so we can't uh put money into such speculative things, and it's a lot harder because there's no funding. So, of course, the UK Innovate UK doesn't invest in US companies. So if we want to collaborate in US companies, there's no real funding opportunities to do that, so it's far more speculative. So I I think that's why we see a lot of uh work being done in the UK and the EU because there's so much grant money um and focus on on these problems, which is an amazing thing because the this sector is really struggling, and uh there's lots of reasons to dislike what the UK and you is doing for agriculture, but to be fair, that part they are doing quite well.

Andrew Jones: 41:44

Well, I mean we we often get Innovate UK mentioned I mean the uh podcast recently on the care recovery bucket. I they got some money from Innovate UK to go forward with it, but I think we've had that. What have we done? We've um obviously spoken at Southwest Dairy Development Centre, we've had um OxyTech on, Herdvision on Hoof Count, all came back to we I remember Anthony Marsh making that comment with his Hoof Count product, uh the Pettiview, he said I would have done it anyway, but he said it probably would take me five years and cost me a lot more money. But he said, thanks to Innovate, I've got money a grant to do it, but also it gave me the links to talk to the right people to bring the right resources in, so it happened a lot quicker.

Chris Knight: 42:28

Yeah, and I think uh to broaden that out a little bit, just from a tech, a tech company perspective, normally the the progression is we come up with an idea, you get seed money, which is 50, 60, 70 grand, then you get your uh series A money, which is hundreds of thousands, maybe millions, and then you get your series B, which is tens of millions. But each one of those stages, you're giving away 20% of your company, which is fine and we're completely okay with. But it's not so much the 20% of the value of the company, it's 20% of the decision making within the company. So rather than be, I've been keeping agriba, I've still self-funded it, but I've only been able to get to the size it is because of the grants and keep it uh and and not take on investment. And while bringing it up is that's key because it allows us to be mission-led uh rather than worrying about if we take a million pounds, for instance, an investor wants 10, 15, 20 million pounds back. So that it then becomes that that then becomes a bad part. Can we do this? And you you can do an interesting project, but unless you can 10 or 15x the value put into it, you can't you don't really go for it. Whereas because we're uh self-funded with massive benefit to Innovate UK, we can be more mission-led. So now we can do interesting projects that you know we tried to make them cost neutral because we're a startup, but it we don't have to worry as long as we're covering our costs and it's interesting, we can do it and execute. And I don't have to uh justify that to people who are only wanting a financial return.

Sarah Bolt: 43:56

And from the point of view of farmers on the ground, sorry, from the point of view of farmers on the ground, that's absolutely awesome because you can hear your passion for the industry coming through, and it means that you're then doing something that will benefit those farmers directly. And I think that's that's really amazing.

Chris Knight: 44:12

Yeah, I I think most founders do what they do because they're passionate about it. Otherwise, why would you do it? I would get a job somewhere if I wasn't passionate about it. But where they get lost sometimes is when investors come on board. Um, there's good and bad investors, again, I'm generalizing, but that's when founders lose their passion and their and necessarily not become mission-led anymore because they have obligations to their investors. And it's not that they're compromised or whatnot. That's just the deal, isn't it? And the so from a tech perspective, uh, I think that's the wonderful thing about Innovate UK and um EU money and the money we from we got from EIT is it allows us to be mission-focused and purpose-driven for longer. And then when we do raise, it's a small percent of the company, so we can continue to be so even afterwards.

Andrew Jones: 45:02

Yeah. So taking a step back from from grants and innovate and all that kind of thing, I suppose going back to people on farm, how accurate is it compared to the the good old-fashioned plate meter?

Chris Knight: 45:14

Well, that this is a this is all hoping to come into this one because the the this is this turns into an AI discussion pretty quickly. So when you're doing Earth observation, the satellites, of course, do not measure uh grass growth. They don't measure core from cover from like we discussed early. That they're built generally, they do multiple tasks. We've just in our purpose, that's how we we interpret the data they produce. So we have to turn that the satellites will, for instance, come back with backscatter, which is sometimes decibels, which is a mile miles away from grass. So then that's the AI's job is to turn what the satellite's unit is into uh gi matter mass per hectare, for instance. So when you do that, you have to do what's called ground truthing. And by that we normally mean um measurements on the ground. So we can calibrate what the satellite saver and have a measurement on the ground to say, well, that when the satellite says this, this is what we measure as a ground at the same time. So then we do the modeling that way. The problem historically with uh grassland, and where I think the reason why we'd not seen products for it so much is that it's difficult to correlate what the satellite says with what the farm measurement was, the ground truthing data was. And I, you know, it's tricky because the method of measuring it isn't accurate. I'm not going to point the fingers at farmers saying they're inaccurate. I think I think the measurement itself is inaccurate, to be quite honest.

Sarah Bolt: 46:36

It's a very basic tool, isn't it, realistically? The grow, you know, the plate meter is a stick with a plate on it.

Andrew Jones: 46:42

But you've still got to get it right, because I know I've been on farm before, and maybe people here don't realise, but there are multiple uh formulations on there because places like New Zealand, on a set day within the year, you change that formulation. Whereas here we use the standard formulation, which off the top of my head, I think is 640 by 125. But I mean, I've been on farm before, and their plate meter they hadn't realized. I think they just put it pulled it out of the box and away they went was completely wrong. And you're like, huh, so you've you've got to make sure that's right. And you can have two people follow the same route on the same day, and it's not gonna be a hundred percent accurate, it's a bit like a silage sample. You can take one silage sample, take two bags out of it. It's never gonna be exactly the same. The roots can be different, the way they plot it's different. It's just always gonna be just that slightly different in the way it is. It just needs to be relatively um what's the word close enough? That's not quite the word I'm looking for, but it's got to be relatively right, hasn't it, within within 100, 200 kilos? Because you and I could go, Sarah and I could go out with plate meters, and we'd probably be 200 kilos or so within uh of dry matter per hectare, different in our results. So uh it's all relative, isn't it?

Chris Knight: 47:54

Right. And it you can go further and say, like uh like was hinted at like what are you measuring? You're basically measuring how much the grass pushes back up on that plate, not actually the grass. And then even the unit itself, dry matter mass per hectare, who actually goes out and dries it and removes the so that's an estimation as well. Anyway, I I bring that up because when you talk about accuracy, as well, compared to what? I mean, so that that's the I we have what we our accuracy compared to plate measure readings, but what does that mean? I mean, so we've agreed that that's not accurate, so then how can we measure our accur accuracy against that? So that's why I sort of hinted at earlier what we do is that's why we have a mult multiple opinions. So we have different models, different types of models, the satellite input, the weather input, the farm measurements if we have it. And then part of the model's job is to take mash those up all together and come up with some sort of uh baseline and measurement from it. Um, and in fact, part of the EU Horizon bid we have going in, so that's an 8 million pound project split across 13 people with different work packages. One of those seven work packages is specifically on well, what is the most consistent measurement we can introduce for grass farming? So right now we measure dry matter mass, for instance, because that's what people are used to using and that's what they want from us. But if I was to push back, I I would use a quote from Henry Ford and say, well, if I asked my customers what they'd want, they'd say a faster horse. Because just because that's the industry standard doesn't mean that's the optimal solution. And I don't think dry matter mass, whilst of course we do that, I would do that reluctantly, knowing there must be something better to dare to do it. And if we take that bigger, the bigger issue with that is I think it's holding back grass research a lot because what you can't do from an academic perspective is compare studies across different fields. So you'll do well to uh do comparison studies on different fields, never mind different farms, never mind different regions, and all those kinds of things. So that's this map accuracy and the truth of what's on the grass is actually a massive hindrance.

Sarah Bolt: 50:03

Um you're saying nobody goes out with a pair of scissors and and chops it off and does actual dry matters to treat it as well.

Chris Knight: 50:12

I'm sure there's an academic out there that does that.

Sarah Bolt: 50:16

I'm sure there have been studies doing that in the past. I'm sure there have.

Chris Knight: 50:20

I would question the prioritization of a farmer's doing that. Yeah. Yeah.

Sarah Bolt: 50:25

No, I didn't I didn't know if you'd done some research on that actually looking looking back to or that's what we want to do.

Chris Knight: 50:31

So that but then that's why it's uh a work package specifically for this horizon thing. That's the research part of what we're doing, is looking into how do we just make even the the process of measuring it accurate. Never mind before the talk about comparing um AI and salads versus uh plate measure reading, because there's no real truth between those two. Um it's not a great clear answer, but I'm highlighting part of the issue that we're all we're facing in terms of improving grassland measurements uh generally. The the one the one the one I would say the Rollers Royce I have learned, and we're actually working, uh talking to a company in New Zealand that does this. They've actually got a uh a laser robot that drives around the field with a laser uh measuring uh the whole field. Harder to push back that guy's attitude, uh that guy's accuracy. But uh that's the lengths you kind of have to go to. So that we're really excited to try and do that project because then that really does answer the point, the questions we've just raised on well, what is the truth, you know, in terms of what the what the measurement is grass. So again, when people use our tools and we we put out dry matter mass, again, I would argue these they view that from a relative perspective in terms of what does that mean relative to that field, as opposed to comparing numbers across fields or even farms. Um, you know, you can get two farmers who are really good at eyeballing their own field, but they'll disagree with a farmer on a different field because their eyeball tells them something different. At the end of the day, they just want to know if I put my cows on there, how long before I have to move them again? That's really what you're trying to get get to, not so much arguing about dry matter mass.

Andrew Jones: 52:11

So if you don't think kilos of dry matter per hectare is necessarily the the unit of measurement, any thoughts of what it should be or waiting to see what the outcome of the uh investigation is?

Chris Knight: 52:23

Uh well uh no, I I I think if I knew I wouldn't I could save ourselves a lot of money. Uh but the I I think I I think I hinted at it there. It's more along the lines of like how long can a cow stay in the field, right? Yeah. Uh and be sustenance.

Andrew Jones: 52:42

Or like at least kilos of dry matter, because if you know how many kilos of dry matter over the the area, you know how many cows you can do, a quick fact packet calculation and go, well, that's two hectares, I need it'll last me two grazings or whatever it is.

Chris Knight: 52:59

The issue there is you're accepting you're you're treating uh your dry matter mass as uh as a quantified truth. So that that's the because yeah, you're right. You're 100% right. If you're if you had accurate if the ground if the dry matter mass was a truth value, then yes. So which you could but then I think farmers have internally calibrated their minds on their field and their herd for doing that. So that's kind of why it works, because the farmers worked out in their head that that for that ratio, but you can't take how they've done it to a different farm at all. And that's the that's the point, that's where you get the issues. So I I sometimes uh sometimes I even talk about well, why are we even forecasting um dry matter mass? Maybe we should just forecasting the the milk produced on atmospheric states and miss out because I talked to earlier how we sort of go from atmosphere to weather to uh grass growth. Instead of we go atmosphere to grass growth, why not just go from atmosphere to milk? But I really then I can because that's what we want.

Andrew Jones: 54:02

And then I can say that depends on the quality of the grass that's in there, doesn't it? Because you know, the the I can't remember what the uh where it is. I've got it somewhere, I think it's on the website, but you know, you can take two different bags of grass seed. There might be um only a five or an acre difference in the price, but the difference in value of milk you can produce out of it can be something like it's over a thousand pounds at one point or something, I think it was, depending on where milk price was. It's it's a considerable amount, and yet people worry about that five pounds. Now I know that's easy for me to say when I'm not paying the bill, but actually that return on investment is potentially massive if you use the the the right mix for your farm.

Chris Knight: 54:42

Right, and uh and that's that's again why it's a discussion of what we do with that because you're 100% right, and that's why I didn't jump on seeing uh the answers. Because everything we've talked about so far has been dry matter mass, right? Quantity. But you've just brought up that's not just about quantity, you've also got quality as well, which is a whole separate, different way of measuring it. Uh we we can say you've got 2,200 uh kilos per hectare in your field, but who cares if it's all weeds, right?

Sarah Bolt: 55:09

So that's the So actually, why aren't we why aren't we talking megajoules? Why aren't we, you know, that would be more sensible, wouldn't that?

Chris Knight: 55:16

You know, because actually if we know answers on a postcard, I would say on this one.

Sarah Bolt: 55:21

Because it's the megajoules that's that's driving your your milk.

Andrew Jones: 55:24

But if you're driving megajoules, it depends entirely on what you're cow because each liter of milk is different. We all talk about standard what 5.3, 5.4 megajoules of energy per liter of milk, but that's based on I can't remember, is it four point? I can't remember what it is, the fat and protein, exactly what it is. But if you've got a diff higher fat and higher protein, you're gonna need more energy to produce that liter of milk.

Chris Knight: 55:45

I would argue that that's a that's a good candidate as well.

Sarah Bolt: 55:48

Um but uh that's but you probably know you know what your your milk is, don't you? So you know what that fat and protein is, so you know how much you need, and you know your cows and their sort of feet conversion efficiency. Could be mega joules of energy.

Chris Knight: 56:03

I like it. Uh but I'm sure someone will come up with a reason why not. I'm sure it's not. Uh but it doesn't mean they're sure. Yeah, it could well be. But that that's why we want that's the point of the research, right? Is to ask and really look into it, come up with a with uh with a uh arguable solution. No, I'm not really into finding perfect solutions, but at least an arguable it might be different for different farms though.

Andrew Jones: 56:29

It might be megajoules, it might be kilograms of milk solids per hectare. It depends on the farm, their system, and what they want, maybe.

Sarah Bolt: 56:37

Depends what their motivations are.

Andrew Jones: 56:39

Yes. And it should be easy enough, I would have thought, within the system to go, right, you want this, you want this, Sute. It's just slightly different, but it's the same data just being shown in a slightly different way, isn't it?

Chris Knight: 56:50

Yeap. And then we've got to worry about air forecast.

Andrew Jones: 56:56

But but it I guess I suppose it's uh it's a I just thinking the example of myself, you know, went to Australia, I was used to talking percentages. In Australia, it was kilos of milk solids, and I found that hard to start with, and then it was, I think one of my key meetings said, Oh, can you? Because I used to show them the you know the results, the staff meetings. I used to then convert it into kilos. So now to me, I start thinking in kilos, and and some people here are now thinking in kilos. Um, but it it's sometimes it's for example, I had a client retired from daring now, used to worry that he was at the bottom of the um league table for the supplier he was, or near the bottom, I should say, because he didn't have the high solids. And but it's like, yeah, but in terms of kilos of milk solids, you're probably producing more than anybody else on there or there or thereabouts. You've got the litres, but those guys at the top at the time would have been this time of year, October, November, would have been spring carvers, low yields, high solids, so not producing the kilos in the same way. And once I sort of explained it like that, he went, Oh, no one's explained it like that before, and he never worried about it again. Whereas up to then he's always worried, oh, I'm in the bot 20%, I'm gonna, you know, it's gonna affect me. It's like you're producing the milk solids.

Chris Knight: 58:06

But I think that turns into uh that brings up that a good point, is that as much as we talk about, well, what is the unit or what is the way of measurement work, part of that conversation as well is that we don't is the the the dissemination of that, right? We can come up with the perfect measurement or the one that we should all be using. It doesn't mean that's the one that's accessible and everyone wants to use, right? So I I've just hate I've just hated on uh kilograms per or gi matter per kilogram, right? But if if that's if the markets, if people want to stick to that and it's easier of dissemination, then we have to think our way around all that kind of stuff as well. So I'm not trying to discuss changing it for the sake of changing it. It's just a conversation about well, what is the best measurement, what is the best way of doing it. But that includes people coming along with it.

Andrew Jones: 58:51

Yeah, and going on the journey with you to explore it.

Chris Knight: 58:54

Yeah. So if they just don't want to, if if the if the new way is just too complicated, then it feels then it'll fail that whilst it's accurate, it will fail on the acceptance of it.

Andrew Jones: 59:06

Well, it's the whole beta max VHS thing, isn't it? Is the beta max was technology, technologically better, but the VHS was more usable in terms of the films and that that people wanted to watch?

Chris Knight: 59:17

Yep. Don't get don't get mad. Now I've got lots of examples, and you'll get really into my mind if you're going to the technique.

Andrew Jones: 59:26

But yeah, but that's so okay, we've gone around it, but relative to plate metering, how accurate would you say it is, or how how relative is it to it?

Chris Knight: 59:39

Well, we're we're aiming for 85 to 90 percent, put it that way. But that's the caveat of all what is accuracy.

Andrew Jones: 59:46

So that's the numbers we're gonna do. Sarah and I could go plate meter and get probably that much difference between us anyway. So, you know, you I'd argue you can't worry to me, and tell me if I'm wrong, it's more about it's not necessarily that data. on the day it's more the the what's the word I'm looking for but the overall isn't it the jour the not the journey but the type the longer time so ultimately you want to see that there is growth or there is this or there is that that's probably more where the what matters isn't it rather than necessarily being within that hundred the forecasting that's the the the the bit that they currently haven't got rather than worrying whereas they can go out and measure rather than worrying about whether it's a hundred kilos per difference it's the fact that there is growth that is going forward and the fact that say two different people can get that same result different result anyway.

Chris Knight: 1:00:39

I I think one thing I I try and uh you know and again this is a part of conversation we're having a short a short term conversation is around well is we I think it's more interesting and useful to be consistent than necessarily have an argument about accuracy because if we're consistently giving the same measurement I think that's what I've been trying to say very badly that we're consistent about it people will calibrate and and they'll they'll they'll go better with that rather than arguing about well we said 2200 he said 2000 the the the plate measuring said 1800 we'll be consistent we'll at least be consistent if nothing else and then we can uh leave it to the academics to argue what is what is the actual number it's consistent that that's what I was trying to say very badly is the consistency of the result rather than it varying which can be different person or whatever it happens to be. Yeah yeah yeah yeah yeah so um I'm looking at the time and would you believe it it's time to sort of go hmm so any last words of wisdom from yourself Chris or anything you want to say that we haven't discussed uh no I I I've got to be careful I don't show all my wisdom and how quickly it runs out uh no I think it's been a great conversation uh look there look there is uh but I think for an hour we probably covered it all uh if you were to give me five hours I'm pretty sure I could go on and we could test people's patience see that at what point they all drop off if they haven't already oh dear yes I think somebody once said to me oh I'll use your podcast to go to sleep oh gee thanks well so long as you get the listen count right I get paid either way that's why there's no test at the end oh dear Sarah I think it's just um what I've learnt from the conversation is how AI has got a real role to play um in something that's that's so um grounded in um in in nature and in the industry that actually you know sort of the two ends almost of the spectrum that farming is such an ancient um career an ancient what's the word I'm looking for I can't can't find the word but that you know farming is is such an ancient thing and we're talking about AI and how it's actually able to to help that um that farming process and and everything else.

Sarah Bolt: 1:03:17

I think it's just another example of of how we need to embrace it within our industry.

Chris Knight: 1:03:22

Well I'll give you a bit of an insight I think agriculture sometimes beats itself up and not what you said but in a in a set regard is that when we're hiring AI people is really difficult. And by AI people I mean like people who understand the nuts and bolts of it not using ChatGPT um although it's valuing that but it's not an AI engineer let's say it that way. But when we when all the good ones like we we have really amazing uh AI guys working on it because agriculture is actually a really easy way of luring them out of corporates and into the real world because everyone I think these days everyone thinks of AI as chat GPT and all these kind of things. But yeah but the reality the reality is like any 95% of the useful AI models will run on a calculator uh because large language models are big because it's they are what they are. But AI is all mainly all these little models and AI engineers want to work on those not do some API integration into ChatGPT so when you're talking about people come and join our AI when I spend five minutes and tell them what we're working on and how we're doing it that they'll drop they'll drop Microsoft they'll drop Facebook in an instant because of what you just said it's a real problem it's impact and it's something they want to work on. And agriculture is actually a really good tool of luring the best tech talent into it if done right because it's a real problem and it's a primary sector.

Sarah Bolt: 1:04:53

And as as as an industry we we really need to embrace that don't we and and make sure that we can uh can get the most of that and that for both for both both the AI techs and for and for farming but I would say agriculture is actually not that bad.

Chris Knight: 1:05:08

I I think farmers get a bad rap on that I I would argue education's far less uh developed tech wise and just not wanting any of it they all talk about it but uh because I I was consulting for a a big education company all the parents will talk about they want their kids to do the latest and greatest and but as soon as you say well we're trialing this AI model well not on my kid you know no no not not my kid I want I want AI in education but not on my kid so that's why education education really stuck uh where it is and I think I yeah agriculture's got its naysayers but every sector does but I would say it's it beats itself up more than it should in that regard as well.

Sarah Bolt: 1:05:48

It's also got its innovators hasn't it those that you know those early adopters that will will get out and try it.

Chris Knight: 1:05:54

I I I would argue agriculture's behind because it's been undervalued for so long. I don't think the that the fact it's behind is its is its fault. I think people have just taken it for granted uh and it's gone under the radar for too long rather than the problem being within agriculture itself.

Andrew Jones: 1:06:11

Yeah there's some real as you say innovators out there there's some people early adopters of things yes you do have people I'm just seeing of someone I know who's got a parlor probably older than me that they're still milking in whereas you've got the flip side of that you got people with robots and uh embracing tools like this or hoof count or whatever it is using different AI models is is there's some real thing. And I think it's sometimes farming does have that I suppose image that's still the country bumpkins and but really these days and and you know it's a bit like when I was at school well I'm not trying to knock in if you were on the thickies you went and did the um you went and did the farm because we had a farm on the on the school sort of thing and yet these days to be in agriculture really you know you've got to be switched on whether it's to use the technology and that you've got to be aware of what's going on. It's not just um you know yes there are some menial tasks still which we've all got to do but there's a lot of technology these days in just driving tractors or whatever it happens to be it's not it's not what it used to be in that regard.

Chris Knight: 1:07:15

No I think you're right because I I would like to own a farm at some point I I am serious about it.

Sarah Bolt: 1:07:21

Me too I'm I'm not uh I'm not from a farming background either Chris and it's it's one of my ambitions.

Chris Knight: 1:07:28

But the yeah but the one thing I'll say is you know I would have to do what Jerry Clark I would have to be rich enough to be really bad at it for five six years because I don't think it's easy it's got this rap of being like country Republicans in the world but I'm close enough to see it and it's just like no I couldn't do it. I'd I'd be awful at it and you know because there's a lot of thought there's a lot of knowledge in there there's a lot uh but you can but then you know again you can be you could go into intel one of the big things about AI is well artificial intelligence well what is intelligence like what is it like I remember I did a talk six years ago seven years ago uh about why Trump was more intelligent than Einstein on the premise on the premise that Einstein would never win a presidential election never do it because see what you want about Trump and what he does with it you can argue with that but he is emotionally intelligent he knows how people react he knows how to get a reaction out of people and he knows what he's doing and that's but that's an intelligence that's underrated whereas like we're and I benefit from it because I'm good at baths I'm good at physics and you know I get labelled intelligent intelligent a lot because of that I'm I'm easy ticks but at the same time it's just like well and I bring this up because farmers are the same thing like but farmers is an underrespected intelligence to take all the variables that they get given and then to come up with a decision at the end of it. Why is that underrated? It is but why it shouldn't be because it feeds us all because intelligence because intelligence is uh hoarded by um maths and physics I think that's the unless you do one of those two it's hard to get that label.

Andrew Jones: 1:09:05

But so wrapping this all up now because we No no no that's the whole point we've gone on and had an interesting conversation on something else. It's been great to listen to you again Chris obviously because I met you earlier this year. I you know I think it's something that's really exciting. I think it's something that will be potentially bring massive benefits to people as you say potentially save half a um a half a day a week you know in terms of your grass measuring it's got some you know really great things moving forward and it's been good for you to share with us and we've obviously had some good conversations and we we talked about again we've talked about innovation within the UK dairy industry as you say it's because of the grants and that available and we've had some wonderful people like yourself on here and it's been an absolute pleasure to listen to them all because there's it just the absolute passion there is and and you know in fact you went hey I'm gonna start helping agriculture and and and then and some of the things you said about trying to bring people into the AI into agriculture um you know you find it easy because people want to help real problems and solve real and you know solve real questions not just doing something for the sake of doing it sort of thing so that that's great to hear as well that you know there is a positive for uh agriculture and it's not always easy to see thank you it's been wonderful to be here as well and have the conversation and uh to give me a platform to rant onto no no no so so on that I'd like to say uh thank you and it's a goodbye from me and it's a goodbye from me goodbye from me thank you very much thank you for listening to the Tune the Cut Podcast podcast for the UK dai industry brought to you from the southwest of England and listened to around the world now for the really boring bit I'm afraid the legal disclaimer the information provided during this podcast has been prepared for general information purposes only and does not constitute advice. The information must not be relied upon for any purpose and no representation or warranty is given to its accuracy, completeness or otherwise. Any reference to other organisations, businesses or products during this podcast are not endorsements or recommendations of Tune the Cud Limited. The views of Andrew Jones are personal and may not be the views of Tune the Cud Ltd and the views of Sarah Bolt are personal and may not be the views of Kingsay Farming and Conservation Ltd and any affiliated companies. For more information on the podcast and details of services offered by Tune the Cud Limited visit www.tunecud.com Thank you and goodbye