SepaIQ Tech Talks: Sepasoft & ATS Global

Speakers: Mark French, Director of Design Consultation, Sepasoft; Dr. Rico Kusber, Lead Consultant Data Analytics & AI, ATS Global

[0:00:04] Mark French: Hi, hello and welcome. My name is Mark French, Director of Design Consultation at Sepasoft, and I’m joined here in Kassel, Germany by Herr Dr. Rico Kusber from ATS Deutschland.

[0:00:15] Rico Kusber: Hi Mark. My name is Rico. I am Lead Consultant for Data Analytics and AI at ATS Global in the German division.

[0:00:24] Mark French: Well, we’re having a discussion about SepaIQ and AI. And Rico, why are we talking about this?

[0:00:32] Rico Kusber: Recently we met and we had a look at your new product, SepaIQ, and we had the chance to put our hands on. And that was interesting, challenging, and exciting. During that session, we came up with the idea: well, we have to wrap it up. We have to talk, discuss, and provide this experience to others. That’s why we are talking, aren’t we?

[0:01:00] Mark French: Yeah, I really appreciated it. Let’s talk a little bit about the context that we’re working in. We’re serving manufacturers, and manufacturers are struggling to use AI to generate value within the organization, to generate transformation, and take the next steps in their digital journey. What are you seeing for the context of AI and MES for manufacturers?

[0:01:29] Rico Kusber: I think that it is a well-known prerequisite that you have to have profound data. This is your input for whatever system—may be AI or anything else. That also is output of other systems, and we have to connect it, we have to integrate it somehow. This is what I see as the context here.

[0:01:50] Mark French: So let’s talk about the tool of SepaIQ. You got to get hands-on when we met in Munich. What are your thoughts about the nature of this product?

[0:02:04] Rico Kusber: As I experienced in the past, Sepasoft provides MES functionality and is really present in this context of manufacturing execution, monitoring it, showing it, and being able to adjust it. SepaIQ—I wondered myself: is it an AI tool, or is it more like a data management tool? Is it something like an extension to an MES? What is it? I was not so clear in the beginning; I was curious, but not so clear. And then the fog vanished a little, for me at least, and I came up with the idea: why not using SepaIQ as a data management, or more than that, data refinement tool? To put it into the middle of those systems—into the middle of MES, AI, and maybe other systems, like a reporting system, for example. Because I got the strong impression in our session that SepaIQ is really handy to put data things together, data streams together, gather different sources, refine things, and at the end prepare the data in a way that it is exactly the required input for another system. This is SepaIQ to me.

[0:03:36] Mark French: Yeah, I think that’s exactly right. I think it’s sometimes challenging to communicate this because everyone wants to be the one to score the final goal. But we need players to make the assist, to make the pass. I feel that SepaIQ is a tool that is making a lot of passes and making plays happen within our data engineering context for MES, for AI, for business intelligence, and other reporting.

[0:04:09] Rico Kusber: Absolutely. So maybe some of our customers might see it as a data engineering tool, others might see it as a data science tool. For me, it’s an interfacing tool as well. This is a good perspective for me, seeing it as an interfacing tool. What you can do with it is you can go to source, extract data from the source, and it doesn’t necessarily need to be only one source. It can be a database, and then you can connect message buses, for example, or connect the output of another system. You don’t have to have everything in a database, but you can connect different kinds of communication technologies and data storage technologies. And then once you extract it, you usually transform it, and this is what you can do easily. It is quite sophisticated or enhanced how you can put together what you really want, put prerequisites and constraints on it, so that you end up only with these parts of the data in that structure that you really want and can pass it on to, let me say, an LLM, for example. And that LLM will appreciate that exact data input, because the more tailored your data is, the more precise your output will be.

[0:05:37] Mark French: I agree, and I think that speaks to one of the main challenges a lot of our customers are running into with AI programs or projects. They’re finding out the hard way, unfortunately often, that cleaning and contextualizing data before it reaches AI is really critical to success. I think this tool is well-situated there. Well, let’s talk a little bit about the ecosystem and some of the neighbors next to SepaIQ. For your part, where do you see this being used among other systems and products?

[0:06:22] Rico Kusber: For the fact that we are system integrators at ATS Global, we integrate anything that is amongst the automation pyramid. It starts with PLCs and machine automation, but it goes up—SCADA systems, MES, and at the end ERP systems. The ecosystem I see here is production data in whatever level of refinement we have. Parts of the ecosystem of SepaIQ are MES, SCADA systems, reporting systems, and business intelligence systems all along the data pipeline.

[0:07:08] Mark French: I agree. And it’s worth pointing out that we at Sepasoft are a premier technology partner with Inductive Automation. So we, of course, want to work with Ignition—just a fantastic platform. But SepaIQ can also work with other platforms, so I see it also as a technology bridge. If an organization, especially large global enterprises, has a diversity of products in their manufacturing IT environment, we can work with others. We’re not only working with one or two systems; we can work with other vendors as well.

[0:07:53] Rico Kusber: This is what I also experienced in our training session. When we think about the ecosystem and all the variety of systems, whenever we connect two systems—and we do this as an integrator—we have an interface in between. All those interfaces need to be taken care of carefully, really. You can use SepaIQ for the interface between SCADA and MES, for example, but you can also use it for the interface between a message bus or communication system and connect it to reporting or the MES or anything else in that ecosystem we mentioned before. So you can use it as a kind of interface refinement and interface structuring entity here for different systems, not only for one. And that was also something I thought about, and I thought, well, that’s nice!

[0:08:54] Mark French: Well, let me shift to talk about AI specifically. It is the hot topic in the industry and has been for some time. Let me talk first about your background with AI. I mean, why am I talking to you as an expert about AI, Rico?

[0:09:15] Rico Kusber: My experience in AI started way back in the days during my studies already specializing on AI, around 22 years ago. AI these days had nothing to do with LLMs; they did not exist. Of course we did have neural networks, which were a great paradigm already, but there was way more beside those. There were decision trees, expert systems, fuzzy logic things, probabilistic systems—all such a large variety. This is actually what I enjoyed. My favorite paradigm was artificial immune systems, can you imagine that? That was great. But AI developed quickly, and these days it’s focused on neural networks and LLMs in particular as a special kind of architecture. Beside that, I will not forget that there are different paradigms that might play a role again 10 years in the future. This is what makes this topic interesting for me, really.

[0:10:36] Mark French: Well, I appreciate that. That kind of history and knowledge of the development of the practice of the art is very important. One of the things system integrators bring to customers is your expertise and experience, so you understand why things are done the way they are done. As we say, it’s not your first rodeo! Thank you. I’m wondering for you, where do you see value and where do you see challenges for AI use for your manufacturing customers?

[0:11:19] Rico Kusber: Value I see a lot up and down the automation pyramid. There are a variety of use cases depending on which level you look—shop floor user interactions, quality management, predictive and prescriptive maintenance. Everywhere there is use, in my opinion. But there are also the challenges, and these are mainly concerned with data. We learned already that the input data you use is crucial; what you get out is only as good as what you put in. But that’s easier said than done to refine the data in a proper way. First it starts with understanding that the data is connected, that data coming from different sources has different properties, and that you need a certain kind of data quality for this or that system to be a valid input to receive a feasible output. These are a couple of the main challenges I see.

[0:12:42] Mark French: Well, so I’m glad that our universal law of computing, “garbage in, garbage out”, still applies!

[0:12:48] Rico Kusber: It does!

[0:12:49] Mark French: Now, forgive me, in German would I say “Müll in, Müll aus”? Does that still work?

[0:12:54] Rico Kusber: Yes, it works! You would be understood then.

[0:12:58] Mark French: Okay! Well, it’s probably not going to catch on just yet! Let me talk a little bit about MES data in particular within SepaIQ. We’ve talked about reporting enablement and context, and I think context is king here. The example I like to use is that of temperature. If I have temperature in isolation, that is not very meaningful. Is that temperature from a vessel that is empty, full of product, or maybe in a CIP cycle? Those three scenarios happen every day and completely change the meaning of that data. So that’s just one example. We have a million such examples from the plant floor. That underlines the importance of applying context to derive meaning, which is necessary before it gets to AI.

[0:14:06] Rico Kusber: Yeah, absolutely. When we talk about this, who do you think will use the data and results of what SepaIQ is able to achieve?

[0:14:22] Mark French: That’s a great question. Whenever we have this data, it’s very tempting to only think in one direction and say we will send it to the AI in the cloud or a local model owned by the factory. But we really need to think in all directions. We can provide contextualized, high-value, high-performance calculations and predictions back to the plant floor, to business systems, or predictive activity locally on SepaIQ or externally in large AI projects. We need to think in all directions and consider all those different consumers of data.

[0:15:16] Rico Kusber: Yeah, this is a nice perspective. And there are a lot within a company.

[0:15:23] Mark French: Well, I’m wondering kind of to wrap up, Rico, can you help us understand how ATS leverages these technologies on behalf of your customers?

[0:15:37] Rico Kusber: Yeah, in my point of view, as I said, there are so many different systems within the context of one company. Usually, the customers know their pain points already—that’s obvious—and they want to put systems into place that relieve the pain. But those systems are rarely out-of-the-shelf. As integrators, we put together different systems, and that one helps here, and that one helps there. The benefit really arises when you put them together. What I see there is whenever we put systems together, at least two of them, we have an interface—at least one point in the ecosystem where we have to think about how communication works, how data flows, how things connect, what we need, and how we put things together. This is where I see we could really make use out of SepaIQ. It could really help our customers solve some of their problems, ease some of their processes, or refine the quality of some of their results. Putting together the systems and refining in between what we have as input and as output—because the better we do that, the more precise and usable the output of the whole system is.

[0:17:17] Mark French: Excellent. Well, we’re available for helping customers realize their business goals and objectives through the use of their manufacturing data, system connectivity, and process transformation through digitization. If you’re interested in learning more, we’ll have resources and links below for both ATS and Sepasoft. Thanks for tuning in for this discussion!