David Pelsoci, managing partner at Captios Partners, discusses digital transformation in feed manufacturing, emphasizing the importance of foundational work before implementing AI solutions. He explains how digital twins can reduce logistics costs by up to 25% and increase yields by 15% through scenario planning and optimization. Pelsoci addresses the challenge of system integration, noting that operators must ensure their ERP systems accurately reflect business processes and generate quality data before pursuing advanced AI tools. He outlines the difference between traditional and predictive maintenance approaches, stressing that accessible, accurate data is crucial for success. The conversation concludes with guidance on measuring technology ROI through metrics like cost per unit and throughput.
Transcript of interview with David Pelsoci, managing partner, Captios Partners
Jackie Roembke, editor-in-chief, WATT's Feed Brands: Hi everyone, welcome to Feed Strategy Chat. I'm your host, Jackie Roembke, editor-in-chief of WATT's Feed Brands.
This edition of Feed Strategy Chat is brought to you by the Feed Mill of the Future Conference. The half-day conference held on January 27 at IPPE brings together feed industry experts to examine emerging technologies poised to shape animal feed manufacturing. The event is produced by Feed Strategy and Feed & Grain and organized in partnership with the American Feed Industry Association. For additional details and to view the full agenda, visit www.FeedMilloftheFuture.com.
Today we're joined by David Pelsoci, managing partner with Captios Partners, a consultancy that supports clients through the entire digital product cycle. He's here to talk a little bit about that digital product cycle and what feed mills can do to streamline operations.
Hi David. How are you today?
David Pelcosi, managing partner, Captios Partners: I'm doing well. Thanks for having me on.
Roembke: Thanks for being here. Well, let's get right in to it. What are feed mill operators getting wrong when they think about AI and digital transformation technologies?
Pelsoci: It's a good question. The easy goal and target is to go straight to all the exciting stuff and skip over a lot of the key sort of foundational building blocks that you really have to make sure are implemented. So we actually, we kind of joke with a lot of our customers and call it the eating-your-vegetable phase.
A lot of times that shows up in the way of you still have to do things and think about process improvement, you still have to think about data quality, those things are very challenging things to fix, and one that you can't wave a magic wand and hope that it's fixed without the hard work.
Another thing is thinking about your overall tech capability and budgeting and planning for it. Oftentimes we work with early customers who think it's kind of a one-off and a one-and-done project. I would actually think more of your technology and digital data software, AI as an asset and that asset requires maintenance. It requires reinvestment in order to get the benefit and value that you'd expect out of it. The last thing I would say on that is really trying to take on a lot of these more ambitious technology projects on your own. I can't tell you how many customers we began working with, where it began as a side project.
That ultimately doesn't play out very well. And the challenge there is that. You lose a lot of trust and momentum that is very important to get with a lot of the tactical people on the ground that are living this day to day. So it should be taken a lot more formally and seriously and we would say at a minimum work with a partner that can at least give you, be a sort of river guide and give you a North Star on what you should be thinking about and how you should go about it.
Roembke: In a previous discussion, we talked about feed mill digital twins. Can you please provide a concrete example of how digital twins have helped one of your clients reduce costs per ton, improving throughput.
Pelsoci: Yeah, absolutely. Maybe before I jump in, I'll just give a very high level to recap what a, what a digital twin is. A digital twin is basically creating a digital replica of your entire, in this case, animal feed operation. So that's taking all of the different data inputs. That's ERP data, that's PLC data from your automation tools, that's CRM data, that's shop floor data, inventory, you name it, the whole bit.
And the idea is that you get a digital replica of your operation and this tool can operate as that control center that gives you a heads up, and really enables a number of things. One can give you a heads up of things before they happen in a predictive way. It can also help you optimize and think about what if, scenario planning.
I'll jump into the scenario planning in a moment. But as far as kind of clear examples of where this has really improved from a commercial and cost perspective is when we think about the logistics and the logistics flows area in the animal feed operation, so optimizing delivery routes, load scheduling, inventory optimization.
We've seen up to 25% cost reduction in more complex logistics operations where there's, let's say multiple plants, long-distance hauls, back routes and back haul opportunities. Trucks, as we know, aren't cheap and drivers aren't exactly growing on trees. So optimizing for this is can be really beneficial, and a digital twin is more holistic.
But this is one area where it really kind of pinpoints clear value. Back to that scenario analysis because it is a little bit more broad and covers your entire operation, it enables you to do scenario analysis. So let's imagine, for example, you can ask questions like, if I upsize my mixer, what will be the overall impact on throughput? Where's the next bottleneck? You can model out those changes to reduce the risk of implementation before actually doing it. And avoid costly mistakes and then, and what that enables is really clear speed, improvement of speed in your decision-making because you've got a lot more data and data in front of you as opposed to operating off gut and intuition.
Just one last hard example on that is, one of our customers using a simulation environment that in the digital twin and determining what they were looking at was the effect of batch changes on the overall production and yield. And so using the tool optimized that batching strategy that increased yields by about 15% without any additional investment, which is pretty, you're really squeezing a lot more out with the same amount of resources. And these sorts of tools can pay for themselves, as you can imagine, very quickly, like in that three- to six-month time frame. Lots of hard examples that are very real even today.
Roembke: Well, thank you for that insight and the example. Feed mill operators still struggle with basic system integration. How do you tell someone who's in that position that they should be thinking about using AI in these digital technologies?
Pelsoci: So ERP systems integration very much falls into that vegetable-eating category that I mentioned above. The big question there that we work with our customers and clients to think about is, does your physical real-life business process, is that accurately reflected in your software? Is your software all properly integrated, and does your software and throughout these processes force high quality data in all the transactions that it creates?
If the answer to that is all yes, which is a long journey often for many customers to get to, you're now ready to kind of go to the next stage with the idea being that you now have data that is reflective of your business and you can start implementing more advanced tools like analytics tools to get real time pinpoint visibility into what's happening to your business. Again, no intuition at this point, but all real-time data about what is actually happening at the cost level, at the plant level, at the product level, etc. And it also positions you for now taking on more advanced AI solutions as data, as a key input to build any sort of custom AI solution.
So that's kind of the path that you still need to continue down is to keep you doing that vegetable eating until you get there. However, in parallel, there are options to implement off-the-shelf tools that aren't really necessarily related to your ERP, but can help your improve your day-to-day, I'll call it efficiency.
So examples like integrating a secure connection with a large language model and using that to improve your day-to-day tasks, like all the cool things that large language models can do, but especially in the HR and back office activities, there's a ton that can be done with off-the-shelf tools.
And then, lastly, what we're seeing more of is that ERP and related tools are actually embedding AI into their tools. So you don't have to do it. You get the benefit of their investment, and really get the outcome of it without paying for it upfront, you know? Now I will say those that can find that cost can come through in your subscription model, of course. But, you don't have to make huge capital investments to make this happen. So examples of that include AP automation to speed up, for example, invoice input and reconciliation as an example. You've got bin sensors that use AR and VR capabilities to tell you inventory levels. You've got surveillance cameras that can auto detect kind of unauthorized personnel. So there's some of these things you can do while you work on your ERP or are sometimes even embedded into the ERP processes. So lots of ways you can creatively start getting going on that without making these huge leaps and huge commitments.
Roembke: Let's talk about predictive maintenance. What's the difference between a mill that's doing maintenance by the book versus one that's using predictive models? What does that look like on the ground?
Pelsoci: Yeah, the way I would think about this question in two parts, one is kind of thinking about the scheduling of the maintenance and the preventative maintenance itself, as well as how does the preventive maintenance capability work itself.
I'll talk about the scheduling example first. So, in the, kind of by-the-book scenario, you've got a maintenance schedule that's very sequenced, very planned, very organized and standardized and documented. The problem with that is it is not necessarily dynamic and follows a more strict schedule.
Typically, when you use tools like digital twins, you can dynamically take into account that there might be production inventory shortages, or scheduling shortages, or outages, or production outages are different. Challenges to the production flow in which you might say, well, we actually don't wanna do preventative maintenance, and if production is down that day, we're gonna have the system automatically give us a notification to reschedule that to a better day or at a better time, as an example.
Moving on to the actual preventative maintenance part, you know how that's actually practiced as you, you input and integrate. Many times retrofit your control automation systems and operation lines with these preventative maintenance tools that actually will take all the data in about the tool and historical production data, input data output data, humidity, temperature, every variable you can practically think of and try and give you an idea of when that particular part is going to fail. And then schedule maintenance enough time so that your production is not affected and you can keep that production line up. So it's really kind of thought, we think about it in two parts, different tools for different things.
Sometimes some tools can do both, but that's really kind of how it's actually happening in the real world nowadays. It's worth noting that all of these AI-related tools, having clear data, accurate, usable data. Accessible data is one of the biggest hurdles that we often see for clients. Once you get past that compartment, which is usually again, falls into that vegetable-eating category, you can it's actually quite straightforward in terms of implementing these types of tools.
Roembke: Now, what's one metric or outcome feed operators should focus on to demonstrate that these technologies are worth the investment or working as intended?
Pelsoci: Yeah, great question. We like to tie things back to any kind of commercial outcome and business outcome with the tech work that's considered so pretty simple at its highest level. So clear ROI, return on investment, of these specific initiatives and what we have a clear understanding of what the cost is gonna be and what the potential upside will be commercially speaking.
So are you going to be looking for savings or are you going to be looking for yield improvement or revenue improvement, whatever that looks like. That's how we use that metric as our North Star to really decide if something's worth the investment. Another sort of, I'll call it soft metric, is that having these conversations — specifically in the animal feed space — is typically speaking, there's been a large underinvestment in digital technology and the way to work around that has been to hire a lot of people kind of papering over a lot of the challenges and creating process to fix the gaps that are that exist.
You have a lot of these companies eventually hit a wall of what that kind of manual process can support. And they kind of reached the ceiling of what's possible. So taking that next step where you can continue your growth without adding significant headcount in the form of improved visibility, in the form of improved efficiency that really justifies and creates that platform for further growth in the future.
And then the last point, if we want specific metrics in addition to ROI and one that we see and talk a lot about, that really ties through to my last point is thinking about cost per unit and throughput in general. I mean, those kind of core metrics really kind of set the scoreboard, historically speaking, and looking into the future when we look at an improved process kind of give you the before and after to tell you if it's actually working or not, and if you're seeing the improvements you're expecting to see. So very much tie how we think about tech to the actual business outcome and that scoreboard that I mentioned.
Roembke: Excellent. Now, if you'd like to hear more from David on the topics discussed in this feed strategy chat, please consider attending the Feed of the Future Conference, visit www.FeedMilloftheFuture.com.
Thanks again, David, and thanks to you for tuning in.
Pelsoci: Thank you.