Join the Insider! Subscribe today to receive our weekly insights
Welcome, you’re listening to the ITOT Insider Podcast. I’m your host David. Subscribe to get the latest insights shaping the world of industry 4.0 and smart manufacturing. Today I’m pleased to welcome Jonathan Weiss. Jonathan has a career spanning roles in both IT as well as OT, just to name a few. He worked at GE Digital as the global lead IOT and smart manufacturing.
went over to Software AG as their vice president for emerging technologies, then went to Amazon Web Services as their global go-to-market leader for industrial manufacturing. He holds and held several advisory board roles and then switched over to become chief revenue officer at Eigen Innovations, where he focused on machine vision. Today, he’s acting as an industry 4.0 expert, and that’s quite some experience.
So John, why don’t we kick things off with a short introduction from yourself? Yeah, sure. I’ll keep it short and sweet here. I’m just kind of a simple guy who loves technology and who loves how things are made. I’m a manufacturing buff, if you want to call it that. I love factories. I love seeing how stuff is made. My favorite show as a kid was how it’s made. And so, you know, throughout my career, I’ve had a lot of fun.
you know, putting technology together with that passion for how things are made in factories. so fast forward through about 11 years of experience. I’ve spent a lot of my career in the food and bev and CPG world as well as automotive manufacturing all through the supplier base. And I’ve done all kinds of cool initiatives around industrial IOT programs, asset maintenance and health initiatives.
And then of course later in my career as it became more popular predictive maintenance and more AI based applications. so, yeah, I’d love to tell you more about it as we talk today. We are a couple of minutes in this podcast and we already covered a lot of the buzzwords. This is going to be an interesting talk. Absolutely. So I would say as a first thing, I’d like to ask you, given your expertise in digital mates,
I think it’s an understatement to say that a lot of things are happening in the manufacturing worlds right now. Yes, we’ve been talking about industry 4.0 for a couple of years. AI is now, I would say that the biggest hype is already a little bit lower. We’ll talk about it later as well. In Europe, in North America, we’ve seen an exodus of a whole lot of manufacturing companies to, I would say, the other side of the world.
We’ve seen the currently the car industry struggling a lot. On the other hands, we also see initiatives to bring investment in the big startups are entering the markets. Why Combinator? And it’s an investor company is investing in manufacturing related startups, et cetera, et cetera. So Joel, introduction, short question. What’s going on?
Well, holy cow, there’s a lot of stuff going on. And we would probably need a four hour podcast to talk about all of it. Don’t worry. Don’t worry, people. We’re not going to do it. Don’t worry. But maybe we’ll do a couple different parts or something like that. but no, in all seriousness, I think if you if you’re monitoring the industry, you see people on LinkedIn talking about certain trends, manufacturers have capacity constraints related to labor shortages and skill gaps.
There’s this push for new AI capabilities that manufacturers are investing in to try to augment some of that. And there’s also this shift to the unified namespace and kind of redefining data lakes and things like that, which is all about connecting enterprise data in order to power AI. And I’ll elaborate on that a little bit as we talk, but I think when you really boil it down, it’s a workforce
problem in many ways that we’re trying to augment either gaps of people who are no longer in the factories or people are there but they don’t have the knowledge that used to be in the plants because the folks who were there for many many years have left and there was no knowledge transfer there was no way to capture that knowledge and I’m not sure everyone understands but you know in a typical factory it’s quite common to have maybe one person
who really knows how to fix a critical asset or super, super important piece of equipment. And if that machine breaks, all of production can stop. And there may be only one person in the entire company’s phone book who can fix it to get the factory going again. so really, yes, companies are investing in technological ways to overcome that challenge. But to me, what’s interesting is understanding why is that a challenge?
What, why do we even have this problem? And so maybe that’s where we start our discussion. Maybe we take a step back and kind of talk a little bit about how did we get here? And then we can expand on, you know, where do we go from here? Does that make sense? Absolutely. Yeah. All right. So maybe I’ll start this a little bit for the, for the history buffs out there who are like me and really like to understand, you know, how, how do we get here and what the heck is going on? To keep it really simple.
We’ll start it maybe back in kind of the 1940s, because I think that’s really where a lot of this stems from, to be honest with you. If you think about even prior to the 1940s and early 1900s, that’s really when globalization started to happen. We had the telephone. People were placing orders via phone to other locations, other geographies, other countries for raw materials and things like this. And even way before that, through history, we had trade routes for spices and silks and things like that, right?
But when you think about the evolution of trade and globalization as it pertains to industrial something quite significant happened in the 1940s and this was World War II, right? So 1914 World War I happens and then shortly thereafter World War II happens. I’m fast forwarding a bit just for the interest of time but something really significant happened.
at least for North American manufacturing. And I’m going to explain how this ultimately impacts European and even APAC manufacturing. coming out of World War II, the United States had quite literally decimated the infrastructure of Germany and Japan, which left the United States in a power position as it came to industrials. So the United States had this essentially a manufacturing Renaissance period or a boom that took place post-war boom.
And I think what a lot of folks don’t understand is that during this time, manufacturing grew incredibly by by 1947. In fact, manufacturing made up a little bit more than 30 % of the GDP of North America of America specifically, excuse me. if you think about what was happening during this time, even though we were in this this post war,
boom, so to speak, of production, it was fueling the economy and tons of jobs were taking place. We were leveraging old infrastructure. Factories weren’t really investing in new technologies. They were more focused on the throughputs and the outputs of production, while the regions that were now laggers were actually rebuilding with state-of-the-art technology at the time.
So Germany was coming back full force in inventing and investing in cutting edge technology for the time. So was Japan. Now fast forward a little bit and you think about something else quite significant that begins to happen, which is closer to 1979 to 1989. 1979, you have basically the peak of manufacturing in America. We have
for as far as employment goes. So you got about 20 million people that are working in manufacturing by 1979. Not quite large. This is really the peak. This is like the epicenter. And although by 1979, manufacturing is starting to decline a bit from a GDP perspective, it’s employing tons of people and production is like through the roof. But 10 years later, 1989, we have
essentially a pivotal point in let’s call it the third industrial revolution, the internet. The internet changed everything. The internet connected people in industrial settings in ways that the telephone really couldn’t do. And so now you had manufacturers who had the ability to not just place orders and do supply chain initiatives across geographies and across the world, but they even had the ability to run different departments out of different regions and different.
countries. And so all of a sudden, engineering and R &D is based in one country. you know, fabrication and raw material sourcing are in two different countries. And so you have this real globalized approach thanks to the internet. And I’m just gonna say one more thing, and I’ll pause for a second, because I know this is long winded. But there’s one other critical, critical piece that the internet introduced in 1989, about that time. And it was the the introduction of new types of jobs.
And this impacted all the regions, North America, EMEA, DOC, APAC, everybody was affected by this where there was an introduction of, let’s call them white collar jobs, insurance, banking, finance, stockbroker, all these kinds of things that didn’t really exist pre-internet became quite mainstream. And also a catalyst was university. All of a sudden people wanted bachelor degrees. This was becoming the new norm, associates degrees, bachelor degrees.
And simultaneously what took place was a cultural shift where if you didn’t pursue a white collar job and you didn’t have a college education, were looked down at. And so naturally people didn’t want to become machinists. They didn’t want to work in factories. They didn’t want to become the G code and PLC people. It was not a fancy job. In fact, was almost a negative.
to work in a blue collar role. And so you had a cultural shift to where at one time, like I mentioned, we had 20 million people doing this in the country and it was, you can make a good living and people were proud to do it by 2023. So last year we went from in the 70s and 80s to having about 20 million people in America working in manufacturing to having about 12 million.
So it’s a massive decrease. And today it only makes up, manufacturing only makes up about 10 to 11 % of our GDP. Right. That’s a crash course in history right there, but I’ll pause. It’s a super interesting history lesson because we, would say typically when we talk about the industry 2.0, industry 3.0, industry 4.0, most of the time people are talking about the technology behind these.
cycles or this eras. What you now did is, besides mentioning the internet, what you now did is actually not talk about the technology part, but how that impacts people, how our job is perceived. So that’s a really interesting insight. if you take this, I would say if you take the things you mentioned about the history and…
We tried to look in, I would say let’s look to the last decade or so. We saw this, I would say ever continuing decline in manufacturing jobs. If I look to the statistics in the US or Europe, if you look to the GDP, it’s not that…
Because sometimes people say like, yeah, there are less jobs, but our plans are becoming more, let’s say optimized or whatever. But I don’t really see that happening as well. What are your thoughts on that? Thoughts on the last decades? Yeah, so I think the last decade has been really interesting. There’s been a lot of talk about the workforce challenges, a lot of talk about adopting technology. And quite frankly, a lot of the topics we talk about today, even this topic right now,
you know, we’ve been talking about it for over a decade. Yeah, because for in many instances, we knew about these challenges and people in the industry knew that it was coming, I’d say the biggest difference in the last 2436 months is that now it’s here. And so if companies can’t just talk about plans to reinvent themselves and to adopt technology, that the time truly is now. And I so I would say, we’re starting to see that over the last, you know,
12, 24, 36 months that companies are investing in this. And there’s also a cultural shift again, back to the people side that I think is quite interesting. And I’ll get into the technology more in a minute. Cause I know a lot of people are interested in the AI and the tech, but it’s really important that you understand this stuff to understand why AI is becoming so important. So the other cultural changes happening over the last, let’s call it three to five years is that vocational schools and trade schools are actually becoming a bit
more popular again with people who are looking to now enter the industrial space. And you may ask yourself why. There’s probably many reasons why, but I’ll give you my thesis. I believe that it’s becoming cool to get into manufacturing again for financial reasons because today you graduate university with so much student debt, especially in America, that you can’t even get a job that pays back your loans.
a white collar job, so to speak, right? so, or eventually you’ll pay back your loans, but you know, it may take until your 40s or you know, late 40s, no exaggeration. Especially if you go into a master’s program, which is in many cases, what it’s going to take to get to a six figure career, which is what people are going after. And of course, I’m making generalizations here. There are outliers, of course, but the difference is,
I think people have realized they can go to a trade school, come out basically debt free because it’s significantly more affordable and you get a pretty good job in a manufacturing world, easily making 60, $70,000 a year and you’re debt free. And so I think culturally people have started to accept that that’s actually not so bad. And in fact, you can have quite a rewarding career because that’s entry level, right? I mean, these manufacturing organizations can provide
tremendous benefits for your career and a lot of upward mobility. And so I think that cultural shift is happening, but something important is in the middle there, which is where a lot of the AI and technological advantages are coming in. And the piece in the middle is that we’ve had this, let’s call it the silver tsunami of all the gray haired, seasoned people who have now retired. The younger generation took too long to get back into the field.
And so they’re gone. And now you have inexperienced people that are coming into the workforce and the seasoned folks aren’t there anymore. And the companies failed to capture all of that tribal knowledge that doesn’t exist in the manuals of the machines. It doesn’t exist in the SOPs. It’s stuff that people just knew because they were on the floor every day fixing stuff and, you know, tweaking this and turning that to make production happen. It’s that one operator who just needs to…
listen to his or her assets and goes like, yeah, there’s a problem. That’s right. That’s right. It almost becomes intuition for people that know how to manipulate the processes and the machines in a way that is oftentimes not documented. And by the way, this is not a mid market or medium size, small medium size manufacturer problem. If you go to the largest manufacturers in the world,
you’d be shocked how many data points they’re capturing with pen and paper still today. yeah. yeah. Absolutely. Absolutely. It’s, yeah, we, we sometimes think that, that, that, that’s, that’s IDs, would say so advanced in our, in our, in our, in our personal life. we, do everything on our smartphone and, and, honestly, the first thing you normally do when you enter a plant is put your smartphone away, especially if it’s, if it’s in an explosion proof zone or something like that. But
First thing you do is you actually put the device, your user, the data is away and you go back to working with your skater system or your DCS, working on pen and paper in many cases. Yeah. Isn’t that the best part of working in the industry is that it actually requires a brain. You can’t chat GPT how to optimize a process, at least not yet, generally speaking, but…
You have to think about what you’re doing. requires critical thinking and it requires in many instances, quite honestly, it requires skills that you don’t even get to develop in university. And the time it takes an operator to get to know plants is years. Yeah. And for the bigger ones, light might easily be 10 years before you really go, before you come to a position where you go like, yeah, now I, now I,
I understand this thing. That’s right. That’s right. And so if we if we now kind of push into the technology part of the discussion, right. So if you think about that, it takes a seasoned operator years, sometimes decades to master a process, master the equipment in a plant. How do they do it? Well, a whole lot of trial and error. Yeah, they fail. They try again. They fail. They figure out what the process parameters are going to be to get
optimal throughputs, optimal first pass yields, right? On the maintenance side, it’s figuring out exactly what to do to reduce mean time to repair all of the KPIs that we care about in the plant. And so if you think about all of the trial and error that goes into it, and now, naturally, there are lots of lessons learned along the way, because you’ve learned what works, what doesn’t work. If you don’t have a way to document that,
and you don’t have a data strategy behind capturing all of those different manual adjustments that have taken place in a variety of enterprise systems, because data is gonna exist in your quality system, it’s gonna exist in your CMMS system to figure out what did maintenance do to resolve a quality problem. It’s going to exist in your time series data and your historians, especially if you’re a process manufacturer to understand process variation and process drift and shift. And most importantly, the corrective actions that were taken.
And so today, where the industry has evolved to is a place where manufacturers are now embracing technology to connect the dots across all of these enterprise systems. And that’s kind of step one, connect the dots across the enterprise systems and get your data into a single platform, a single source of truth that is a standardized approach to data governance. The latest
trend or kind of norm in this would be the unified namespace that Walker Reynolds did, you know, and the team did a great job defining. And what’s been really interesting for the first time in my career, even going back a decade when we were talking about Hadoop and big data systems and stuff like this, I never saw manufacturers asking for it, like by name. They knew they needed something, but they didn’t really know what they needed.
Today is really interesting. When I have discussion with a manufacturer, they ask me, what do you think about the unified namespace? Do you think we should be doing that? And it’s actually really cool to see the progress that the manufacturers have made in understanding the significance of that. But that’s kind of, think, where we’re at today is companies are now finally executing on the vision of data governance and data orchestration. If you want to call it industrial data ops, this is kind of the buzzword, I think, that’s being used for it.
That’s step one to build the foundation. Then step two is to leverage all the cool, sexy AI stuff that everybody talks about on LinkedIn. But the reality is, you know, we’re still a bit a ways away from that for most manufacturers. Yeah. And before we set it to the big black hole of AI, let’s say, it’s still thick whilst the back, back towards say the data foundation.
There are quite some challenges around, I say that comes with building a data foundation. Let’s just ignore legacy protocols for now. There are always solutions to work with that. But I often take a look to what IT did in the last, whatever, decade, two decades, whatever, because IT really became centralized, became data driven.
The way you work with data in the cloud requires centralization, requires governance, et cetera, et cetera. So on the one hand, I would say like, okay, let’s learn from what’s happening in the IT worlds. On the other side, in operations, in Boaty, you typically say like, yeah, that’s very special. We’re different, we work with other systems. So that’s that.
You can’t compare it to other engines. What’s your take on that? on the IT versus OT, on the fancy cloud systems versus the traditional historian? What’s happening in your perspective? Yeah, interesting. I’ll relate it back to people because I think it’s very easy to answer all these questions with technology answers. But the reality is at the end of the day, it’s just about all of these topics are change management challenges.
right? These are all people challenges. And it is very easy to have an oversight in that regard and just focus on tech. But the reality is culturally, there’s a divide between it and ot. I’ll spare everybody the explanation of it ot convergence and all that we this is seven to 10 years old at this point. can go to the blast. can go relates. But if you but if you think about fundamentally why is there a separation there?
culturally at least from the people side. I think it’s important to understand how they think. IT is a cost center that’s focused on reduction of cost in technological infrastructure, generally speaking. And they implement governance and they implement ways to save ultimately, generally speaking. I’m not making generalizations, but if you compare that to the OT world, OT is…
culturally a bit more innovative on trying to think creatively about how to leverage technology to do all of the things that impact the important KPIs in a plant. So let’s just take OEE, right? So you’re gonna look at how can we increase the availability of our equipment? How can we reduce the quality defects? How can we boost overall efficiencies? This kind of stuff.
And so naturally you have a tug of war that takes place because IT is saying, we’re trying to reduce infrastructure, reduce spend and OT is saying, well, if you don’t get me what I need, I’m going to go to Best Buy and plug it in myself on the shop floor. And quite literally that’s what happens. Absolutely. And that’s a whole different world of problems when you start plugging stuff in the OT world without IT support. so this whole bridging the gap between IT and OT becomes really important. But I think to get
back to the heart of the question is what am I seeing in that regards? think generally speaking, the mature organizations at this point have understood that both stakeholders need to come to the table when it comes time to do an innovative project or do some kind of digital transformation. And so now it’s very common to have your ops people and your process engineering people at the table with the IT teams and the IT leadership. And so everybody understands the significance because
The reality is the data that impacts production performance and operational efficiencies, it exists in both worlds. It’s data that comes from, you know, IT infrastructure systems, and it’s also data that’s coming from the OT systems like the MES and the MOMs of the world. And so I guess to maybe keep it a shorter answer, what I’m seeing is both players are at the table today in the mature organizations.
And they’re making educated decisions together to ensure the success of these programs where IT is now being much more supportive of say cloud technologies than they were maybe seven, eight years ago or maybe even five years ago. Yeah, I do agree with the fact that I would say the things we need, the technology we need, the concepts we need to apply this new data of Sting to manufacturing now becomes
reality. From a technological point of view, think we are not really hindered. Yes, it’s still a developing, I would say a developing world. And yes, you will probably need to combine a couple of solutions left and right to come to your final platform, but it is possible.
Yeah, generally speaking, just about any digital transformation effort is going to be an ecosystem play of a multitude of vendors. Everyone from ISVs or independent software vendors that have application specific solutions will bring value. SIs and system integrators and automation companies will come in and help kind of be the glue of those of multiple application specific solutions.
These days, just about everything has a cloud component. So, you you’ll have a hyperscaler in the mix that’s supporting cloud technologies as well. So yeah, from my perspective, it’s certainly an ecosystem play and the ones that will be successful in implementing that are the people who bring IT and OT together very early in the discussion to align on the business drivers and the overall needs of the programs. And we’re almost half an hour in our protocols, but I now need to answer the AI question.
I’d love to dive into this topic as well. We already, or you already, mentioned a couple of points. It’s an ecosystem play. It’s about your data foundation, cetera, et cetera. But I would say with your experience, what is the role of AI in manufacturing? Can it help in keeping jobs local or whatever you want to call it?
Can it help in reaching the shortage or labor shortage or the skill gap or whatever? Or is it just one big hype? What are your thoughts on it? Probably a little bit of all of that actually. So certainly a lot of hype. I’m not going to sit here and BS you and tell you everything you read on LinkedIn is true about AI. It’s not going to solve all of your problems and it’s certainly not going to solve them all day one.
In fact, there’s lots of problems AI cannot solve at all, at least not yet. Now, probably the most important thing to remember about AI is that when it comes to an overall strategy of implementing an AI program, whether it’s for predictive maintenance, whether it’s for asset monitoring, whatever it might be, AI today, now this may change in the future, but today,
AI cannot augment domain expertise. You have to know how to interpret the results. You have to know how to mitigate hallucinations and AI responses. I won’t belabor the point. If you follow me on LinkedIn, you know I’ve been very vocal about this, especially recently. But it’s something that’s really important, especially if you think back to the history of how we got here to begin with. The last thing we want to do is start to alienate even more people who have the domain expertise.
So it’s really important for manufacturers to remember, none of this works if you don’t have proven domain expertise and professionals who know processes inside and out that are creating the models, creating the paradigms within these AI models and reviewing the responses. So that’s number one. The other side of this is understanding kind of conceptually how does AI really help manufacturers and
to answer the question, does it bridge the skills gap? Does it bridge the labor shortage gap? I think there are many applications where AI is beneficial, but I think the key to doing these things is to build an AI strategy around the concept of having the AI serve the operators and having the AI serve the shop floor personnel, not the other way around.
I don’t think the data strategy should revolve around your machine operators and your maintenance people trying to serve an AI model that they have to believe results as gospel, so to speak. At least not today, right? No, and that’s interesting because one of the things we say, current way of training models is still data based. it starts with having…
data available, whether it’s process data or whether it’s in manuals or whatever.
But second problem is that for many other organizations, we try to have as stable as possible process, right? So having a process which doesn’t deviate from a set point too much has been the target of the last how many years? So in general, if the process goes well, we actually have a double problem because our new operators who come in
they aren’t exposed to upsets. So they’re not really exposed to how to deal with an upset on the line. They might not face it for years before it suddenly happens. But secondly, if you then wanna use data to predict whatever you wanna predict, then you also have the data to do that. If you see similar things. Yeah, it’s a good application or use case actually. Like if you take the instance of
process drift or shift is really what you’re talking about, right? In a traditional manner, you would use statistical process control type of tooling. And I think that’s actually a really good use case for AI where you can leverage a model that can be trained on what does good look like in a process. And when you do see drift away from set points or standards, expected higher specs, lower specs, all these things, right?
When you start to drift out of the expected norms in your process, that’s where AI can actually become a very powerful tool when leveraged by a process engineer. The difference here is that you have somebody who does have a certain level of domain expertise, even if they’re junior in their career, they understand the process conceptually and with enough rigor that they can validate results that comes from AI. And I think there’s a key there. The other thing too is that
it’s a huge time savings. And so this gets back into some of the efficiency related KPIs where in traditional statistical process control applications, like you mentioned, this could take days, weeks, sometimes months for a sophisticated process to understand what the heck just happened and why was my batch bad. It could take a long time. And so great AI application where now instead of having the process engineers do the digging and the research and try to find the trends themselves,
Let the model bring trends to the surface and use the expertise, the domain expertise to validate what was surfaced. This to me is a great application actually for AI. And there are companies today that quite literally do what we’re talking about. Yeah. That’s an interesting insight. Are there some other, I would say common pitfalls you want to share around going for? Let’s just go like my CEO sat on a plane next to a guy.
talked about AI, came to the boards and says, here is the new, our new strategy is to do whatever, but it has to do with AI. Okay. So, okay. No. Yeah. There are probably a lot of listeners rolling their eyes because they’re in that situation, right? I hope you are in the, probably they are. So what are, in your opinion, I would say common pitfalls. Yeah. So the,
The first one that comes to mind and perhaps the most obvious is underestimating the time it takes to actually train an effective AI model. And I think there’s some misconceptions around what it actually takes. And I think there’s poor salesmanship that’s happening in the industry too, where people are selling you the promise of, know, hey, get started in 30 minutes, we’ll have your model built. And no, mean, the reality is depending on the process and depending on what you’re doing with AI, it can easily take weeks, months.
to develop a robust production grade model with high enough levels of accuracy. So I would be realistic about how long it’s going to take to train a model to get the right level of results. And I would caution organizations from implementing models in production too early, because once people don’t trust the results on the line or in the plant, the project is really, you’re going to have a difficult time succeeding. I would say
recovering from these initial problems or from initial resistance is indeed extremely hard for the degraded data. Very tough. It’s not impossible, but it is very tough. Some other common pitfalls are starting without having that right data strategy in place. So starting with only small pieces of the data, like for example, if you were to…
go full force in with a large language model or an LLM today to try to do some kind of generative chat bot for maintenance purposes. Well, you better make sure that you’ve got all of your CMMS data integrated with manuals and you know, there’s, it has to know where to go find the right types of data. And you have to, you know, perhaps you look into implementing a rag that’s architected in a way that’s specific to maintenance. Like, you know,
I could elaborate on that for a whole other hour, the reality is, I don’t want to belabor the point, but that overarching data foundation is so critical to the success of the AI models or else you’re going to be served up insights that simply are not valuable. absolutely. And it’s also something people, I would say in general, underestimate. Typically, the problem I
Probably in manufacturing is that data governance is typically not really something which is high on the agenda, data management. Yeah. There is also a big difference in, for example, we’re buying a plant which is built by a contractor. They bring in their way of working or especially the food and bed industry, more industries, but especially the food and bed, know, it’s an industry which
lifts or virtual acquisitions, right? So that’s right. We could go for a standardization approach today, but tomorrow we sell off a part of our business and we acquire another part that then again, where I know different systems, different ways of working. That’s very true. This is why the whole concept of the unified namespace is really important because it implements a standard data structure for asset types, different types of data sets. And that can be extended through
really all of the plants, regardless of the technologies that they have. That’s why it’s, I think, a really successful architecture. Absolutely, Ben. It may be also important to note to the audience that, for example, something like Unified Namespace, it’s not about buying your products. It’s about, it’s an architectural concept. Yeah, that’s right. You can’t go buy Unified Namespace in blue or red. It’s
It is a strategic approach to data governance and data management. And maybe, you know, if we want to call it industrial data ops, because I think that is the now the popular phrase for it. But that’s right. It’s a concept and it’s a best practice on how to manage data. So we already had the take enough time. Don’t rush into production too fast. You mentioned the data foundation, UNS, data governance, management. Is there another call up pitfall?
I guess the last thing I would say is not having a true business case behind the implementation that you’re doing. If you’re looking to scale out an AI initiative, maybe this is an obvious one, but I still see it overlooked all the time. You have to have an aligned business.
case and business value proposition and an understood ROI before you’re ever going to get the needed executive sponsorship to actually implement at scale. You might be able to get a pilot kicked off, right? You might be able to do a little science project, but the reality is if there’s no direct business driver and value for the organization at a business level, I would not expect to see success in your program. Can we just touch the ROI points a bit more here?
It’s an interesting topic I just lost week at a discussion where they group about calculating an ROI to invest in the data platform or to invest in UNS alike concepts. And my personal opinion here is that it’s a bit of a chicken and the egg problem. Cause on the one hand you need a foundation with
I yet have to come across the first person who is able to build an ROI on a foundational concept, right? The ROI typically lives at use case level. So if you today, if your customer of yours or whoever a company asks you like, okay, John help us in getting this thing funded. How should a company approach the ROI question when it comes to working with data?
Yeah, so really, this is a really good point in the comment that you made around being a bit of a chicken and egg dynamic. This has forever been the challenge in any kind of data governance or data strategy approach. And oftentimes, it’s thought of, I would say, almost similar to maybe like security in the IoT world, or even the OT world, quite frankly.
A lot of organizations don’t proactively invest in security. They wait until they had a catastrophic data breach or somebody hacked their system and production got shut down and now security is a priority, right? Data governance is kind of the same way. Organizations tend to wait until they have many programs kicked off at once and they realize just how much of a mess the data is. And so all of the programs are failing because there’s no consistencies and governance around how to retrieve data sets, how to calculate KPIs across plants.
And so then they say, okay, we should probably take care of that. The simple answer is one, if you wait till then, it’s gonna cost you way more, because now you have to rework the plumbing, so to speak, on all of your applications. But the other piece of this too is that, to be honest with you, the most successful organizations that go after this AI journey and really embrace new technologies, they…
they understand at the C-suite and at the executive level that this is a necessary requirement to be successful. If you don’t do it, it’s not necessarily that your project is not gonna be successful, but the reality is you’re gonna end up with way more IT debt and way more cost associated with implementing these programs. And so yeah, you have to invest a little bit upfront, but it’s a necessary evil to ensure success in the longterm. And that’s kind of how I would.
how I would phrase it, and I would actually even take it a step further. It’s not just a necessary evil. It quite literally puts you in a better position technically to continue to innovate and continue to do the important things that are related to AI, especially like AI ML ops for model development, model retraining, continuous improvement of models. None of that stuff works if you don’t have your data in order. I think that’s a very…
interesting insights to end this podcast. We did a very interesting talk. We talked about state of the manufacturing, a bit of a history lesson, very interesting one, and another technical one. And I liked it a lot. I liked it a lot, Joe. Thanks for that. And I guess the common pitfalls you mentioned is definitely something people should really be aware of them. And maybe even, I would say,
actively communicate them within the company, within the management teams. Okay, these are common pitfalls. Let’s not make the same mistake others already made for us, right? Yeah, that’s right. That’s right. It just takes a little bit of discipline, but it’s not hard to avoid them. It just requires a little bit of strategy and a little bit of discipline. John, thank you very much for joining this conversation to our listeners. You can find John
on LinkedIn and I’ll put a link in the show notes. Make sure to subscribe and we’ll see or hear each other soon in the app, one of our future ITO.T insider podcasts. So thanks for joining and until the next time, take care. Bye bye.