Vivek Vishwakarma: (0:04) OK, perfect. Sorry. I thank you everyone. I would like to this is the first ever podcast or third AI automation decoding the data loss in industrial automation. And I'm actually very, very excited about today's conversation because our guest is someone I have known and respected over the Kate Stefan, thank you so much for being here. Stefan is a seasoned leader and engineer at heart. He has spent much of his career at the intersection of industrial automation, resilient system and energy innovation. He's currently ASCE of ice, wind and Ireland based wind turbine company designing sustainable energy solution for some of the harshest environment on Earth. Stephen is also a member and awardee of narrow Diana. And he brings a unique blend of engineering, trapped leadership and a real world system thinking. And I'm looking forward to driving into this insight into industrial data loss, RCA and the future of automation. So Stephen, welcome here. And would be happy to. You know, learn from your experience and maybe talk about, you know, the impact of data loss that you see across industry or like you know within your ecosystem where you are active, so it will be great.

Stephen Drake: (1:25) Yeah, absolutely. It's it's a pleasure to be here. Thank you for inviting me on the inaugural podcast. It's an honor. I feel like I've got, you know, some unique backgrounds. So anything I can do to shed some light on my experiences and just kind of what I'm seeing on the industry.

Vivek Vishwakarma: (1:40) Yeah, absolutely. Absolutely so. When we talk about data loss in industrial system. What does that really mean to you? Is it just missing records or does it go deeper?

Stephen Drake: (1:55) It actually covers a couple different basis. So from the idea of like data loss, I see it as two things. I see it as data corruption and data integrity. So you know one speaking to the data corruption is not nefarious, you know, just when you have sensor errors or the wrong types of communication to where you're just not getting, whether it's the resolution or the frequency of what you're expecting. Week I've seen that before, but then also the data integrity. So I've seen this quite a bit. At a previous role where I was leading a global testing program for certifications, approvals and standards. And yeah, we have multiple different types of testing setups. You have lots of different equipment, a lot of different hardware, and you're recording your your acquiring data. It's important to actually take the data yourself, but also when you transmit it to these testing agencies, these etls. You you can confirm that the data's real. You know 'cause essentially from all these different type of data logs you're getting CSE files you know or excel files. You need to be able to say I didn't just enter these in myself, they needed to have a way of validating what the data is and so a.

Vivek Vishwakarma: (3:00) OK.

Stephen Drake: (3:09) Lot of these systems will have things called signatures where they are typically 5 or 7 digits, letters, numbers, some combination of that one, and then these agencies can actually verify the data through known softwares. You know, to make sure that the numbers equate to what you're giving them. So and data losses, I've seen these copy and paste errors intentionally unintentionally, but be able to go back and verify the data is real without corruption is very important.

Vivek Vishwakarma: (3:37) Right. And I know you have been in industry for almost 2 decades. I've known you since 2012, right? Right. And if I'm not wrong, you've been in Navy for some time. You've been.

Stephen Drake: (3:47) 2012 yes.

Vivek Vishwakarma: (3:56) Even before that you you were active, right? So you have basically seen. A very early adoption of Industry 4.0 as well and you know post industry 4.0 adoption, so.

Stephen Drake: (4:07) Yes.

Vivek Vishwakarma: (4:13) What do you think? Like, what's your experience? That how that landscape has changed over the course of two decades? And do you see, even with so much advance hardware, even advanced softwares, we we did a as good job of preserving that data loss or we still see the same same same thing? And then we can go later and digging. Me too. How we can mitigate that?

Stephen Drake: (4:45) Yeah. So just just to clarify in the background is when I was still in university, I did. I worked for a couple different companies, but then my first real time job was for in the defense industry. So it wasn't directly related to the Navy, but he was supplying hardwares and systems and system integration for different departments of the military. But yeah, those were all different areas in in that context, it was a little bit of engineering and as well as manufacturing, so. Similarly, there was a lot of data that was involved there. Then I worked for an oil and gas manufacturing company and recently in renewables. But what I've seen across all those, you know, these company sizes are anywhere from 20 to 30 employees to 1000 employees to 10,000 employees. You know, so small, medium and large. And Jasseen granted anecdotally from the sites that I was I was visiting, and there's probably been 10 or 20 of them of how they approach. Data acquisition. Data processing, data storage, everything. With those and what I've seen is, I think there's the challenges. There's a lot of opportunity because there's still a lot of these companies, small, big and in the medium is that they're not really doing things the way that we think that they're doing it, they're still kind. Of doing it the old antiquated way that we've done it for 10 or 15 years ago. So while there's all these nice tools. And and ability to actually analyze this data. A lot of this is just take the data as much as you can, whatever it is captured, and then the ideas eventually will process it. Later, we'll organize it. Later, we'll we'll properly label it later. But you don't. You just have it. You just you want the data, whatever it is, but you're not actually taking the right steps on the front end to be able to make it worthwhile and useful.

Vivek Vishwakarma: (6:26) All right. Right.

Stephen Drake: (6:35) And and again, I've seen that on, on every scale of businesses. And I think, you know, unless you're one of like the top top tier businesses.

Vivek Vishwakarma: (6:42) Right.

Stephen Drake: (6:43) You know the companies, the people that are actually innovating, all this stuff were very, very small company. That's very fast and nimble and you can do things at at Lightspeed. I would say 95% of the companies out there are just are not doing it the way they should be doing it. You know, from an engineering point of view, because they're all under this pressure to do something quickly. You know whether that is new product develop.

Vivek Vishwakarma: (6:58) Correct, correct. OK.

Stephen Drake: (7:06) Integration product releases, iterations. They're just going so fast. They're just. They're literally just. Is drinking from a fire hydrant to get all the data we'll worry about later. And then if they need it later, they'll worry about that then. But you know it's it's. Hurry up. Do the test without properly going through the right steps.

Vivek Vishwakarma: (7:24) Right, right. So irrespective of the scale and also irrespective of whichever vertical, the problem is very, very agnostic, right? It just boils down how the workforce see or handle the data, even though it's coming from the very, very smartest piece of equipment or how it's coming from. How much you know average equipment, OK. So one thing that I also want to add is.

Stephen Drake: (7:50) Yeah.

Vivek Vishwakarma: (7:53) It's actually the fun fact I was that I came across that engineers are average employee is gonna spend about two hours every day just looking for the data. And doing nothing. So if we're talking about a span of 30 years of career, a person have literally spent 15,000 hours with translate to about 7.5 years of that 30 year career just looking for information. I feel like the precursor to that is a poor data handling at the 1st place, right? So some fun fact, and I feel like this is this is one big problem and. How the data? I mean, this is just one big problem that is kind of, you know. Defines the entire system. Troubleshooting or slows down the entire system troubleshooting. So this is this is this is really good insight. The other thing that I was thinking. Is. Other than the human factors, right? What are the other places where these data loss is happening? Is it happening because of lack of technology? Because of the team or because of the legacy infrastructure. You mentioned that you worked at oil and gas as well, and now you work in a much smaller, you know, company your own startup so.

Stephen Drake: (9:33) Yes.

Vivek Vishwakarma: (9:36) Would be happy to, you know, get some insight on those.

Stephen Drake: (9:42) Yeah. So I I can see from a couple different areas. Again, I I think the part of the reason why it exists is because there's this rush to just do something quicker. And so there is. You have to show progress and the the quickest way to do progress is just to start testing things. Start evaluating things. But the reason why it was to your point, yes, I I totally agree. You know, some days it feels like I spend 30 minutes searching for data or apdf or a document. Whatever the case is. And some days it feels like 4 hours. So 2 hours a day on average. It's it's unfortunate, it's it's suppressing to hear, but it feels right. That feels like the right number, but the reason for all this? I can see it a couple different ways. One goes back to the the original purpose of saying you hurry up, you take data, you're not properly setting up.

Vivek Vishwakarma: (10:25) And.

Stephen Drake: (10:28) Whether it's file names, it's it's headers of of data acquisition. So that. That's a bad start. But then the other one goes to your thought of is it legacy infrastructure? And this goes back to the way corporations are are structured most. Centered around their IT policies. So it policies are similar to certifications and approvals for for products. Their first goal is security. I don't care if it works, it just can't do anything. Negative. It can't do anything badly. And so there are limitations in.

Vivek Vishwakarma: (10:58) Mm hmm.

Stephen Drake: (11:01) Across different size organizations. About different departments, you know, whether it's marketing, its finance, its engineering, its manufacturing, they have all these different siloed gaps. So you may have data, but the way in which you can either. Have data share data source data. Alter data is is most likely different in each different department. You know the size and the scope, so you know. Luckily you have CAD files. You know, there's this some of these CAD files can be absolutely huge. And there's just limitations for these smaller companies. I'm I'm saying anything less than like 1000 employees on how you can, how and where you can store the data. Is it on the cloud? Is it locally? Is it on your laptop? I hope not. I hope it's on a central server and so all these things come together of just this is the way I have to do it for this unique situation, which means there's no uniformity. There's no standard of how it's done. So some person that was doing CAD or drawings five years ago. They probably had a different system limitation. You know the maximum upload size was. Was two gigs, but now it's this tinned gigs. And so there are different locations, different file names. So it's it's a. It's a combination of doing things very, very quickly just to get going, but it's also the limitations of the local IT infrastructure.

Vivek Vishwakarma: (12:20) Correct, correct, correct. And that makes life even actually more, more difficult. Then I'll at this point I do want to, you know, connect a dot and you know, appointed where we are kind of headed. So here automation we have been just doing this massive brainstorming. Like, why do we even care about knowledge loss? What it's doing, we focus mainly on troubleshooting system troubleshooting. So all these knowledge laws we kind of, you know, backtrack to what we just discussed, like a lot of inefficiency of the system and, you know, the person or workforce. And we don't blame that, right. In last few decades, even though we have came a far right, our industries are much, much smarter, but system troubleshooting still has not changed. I mean, we still talk about fish bone. That does not seem right. When you have, you know, hundreds and thousands of parameter that can be a potential model to the failure. And the only reason we found it why we can pin it to is #1 poor reporting or infrastructure or poor practices. That's that's where. You're not putting enough information. To start with, and you're putting information to two many different. Places to start with, so you already lost some of the knowledge. Knowledge is not just. Preserved in a piece of documents, the other thing that we felt is. Irrespective of whatever template you can make when a person is in the system, they have taken so many decisions. If I have a breakfast in the morning by night, I don't know what I had for my breakfast and engineer who is in the environment 8 hours taking you. Know hundreds of decision. To fix a particular problem, how could you imagine them to put every single step you know so that so that if so, that we figure is?

Stephen Drake: (14:13) Correct.

Vivek Vishwakarma: (14:19) It's crazy because the person. The level at which all these industries operate complexity is also very, very high. If some issues like this happens again in future, a person has to do the whole damn thing again. Right. And this will lose a lot of money, you know? I was just going through some of the data and stats. It's posted in our blog as well in our website like automobile loses about two to $3,000,000 every hour. There's a downtime. Right. Fortune 500 companies are expected to lose $1.5 trillion. And that includes all the fixing and delays and all. And that is these knowledge laws is a precursor to all that. You're adding extra hours, you know extra hours you're taking to fix those. The penalty you're paying is just. Way, way too much so.

Stephen Drake: (15:22) And it's not. It's not just the downtime, too. I mean the downtime, obviously, if things aren't moving or operating, but then you're also assuming people are going, people and equipment are going to be 100% efficient when they're working. And that's, you know, just not the case. You know, you mentioned that, you know, we spent a lot of time looking for documents.

Vivek Vishwakarma: (15:35) Exactly.

Stephen Drake: (15:39) That's that's like modified downtime. You know, if you if you work 8 or 9 hours a day and you're searching for documents, then you're kind of not working. And so that.

Vivek Vishwakarma: (15:47) Exactly.

Stephen Drake: (15:48) That downtime is really expanded to more hours of the day. A more inefficiency, so maybe in a given work day, a person is only working five hours a day, no in factor in lunch and and coffee breaks and just looking for things not actually getting the work done.

Vivek Vishwakarma: (15:59) Correct, correct.

Stephen Drake: (16:06) You know you're you're not. You're not getting a lot of practical work accomplished every day.

Vivek Vishwakarma: (16:09) Exactly. Right, right. So this downtime is both active and passive side, right? The active side when the tool is literally breaking but the passive side is all the extra hours been added. With this, you know inefficiency to it. Now this is this is this is great. Have you seen? What are the measures industries are taking to handle it, or at least trying it? We are living in a Gen. AI age so. I would love to see your perspective if you have seen something even at ice wind. If you guys are using anything or planning on using anything.

Stephen Drake: (16:56) Yeah. So the the way I've seen it approach is, you know, companies realize that this is a problem. They realize that when they have, you know, their paying employees 40 hours a week, for instance, they're not getting 40 hours of of practical, useful help out of them. So they understand that there are limitations and efficiency and and they're trying to do what they can to make a difference. You know, whether it's it's a 5% difference, a 1% difference. It's gonna be better than what they're doing. And so they're investing a lot of efforts into. It starts off with software. Whether it's internal ERP systems or Crms depending which which different departments you are just to make, make the workflow simpler. You know whether it's you have a system that goes from 5 clicks to two clicks to do what your system is that helps and shavings make a P.

Vivek Vishwakarma: (17:39) Mm hmm.

Stephen Drake: (17:46) And maybe you can actually work 20 minutes more a week and over a year. That's a whole lot more. For us internally, we're we're still a small, pretty team, but we are we're trying to make small improvements too. We know that. That's using more connected systems. You know the Google suite of of systems where you know how we actually store some some of our non critical documents customer list and as well as integrating you know practical CRM tools is helpful. Using using tools without customization is also nice. You know, a lot of companies in the market offer platforms and solutions, but to make them really useful, you have to do a lot of customization which takes time, which makes them more not standard.

Vivek Vishwakarma: (18:25) Mm hmm.

Stephen Drake: (18:27) And and non standard things across tools and platforms, whether it's protocols or file names just can slow you down. South. Sticking to the basics of the standard approach is very helpful.

Vivek Vishwakarma: (18:39) Right, right, right. No, I I agree. I agree. And I was actually going through one of those reports that 2025 is was the year of implementation. So there was this massive survey that was done. And so I'm. I'm really hopeful that enterprise will, you know, focus more on adopting the external solution and streamline all these data management issues because. This is the time earlier adopt for the eco. One thing that I also wanted to touch in is. Again, since we have seen many, many industries, how does the troubleshooting looks like in your in your book? Is it a very automatic way? Specifically, if I'm talking about, you know, since you have not just seen manufacturing, you were at thermion, which is. From the that made cables right? Like, you know, cables were industrial purposes or telecoms.

Stephen Drake: (19:47) Yeah. Mm hmm. Yeah, is a process heating, but you know, I feel like it's it's it's a very standard engineering problem.

Vivek Vishwakarma: (19:49) And.

Stephen Drake: (19:55) You know when you have lost data? Or frustrating levels of the things that went wrong. It's the great unknown. It's the unknown unknowns and so a lot of times you you really don't know where to start. If this is a problem that's come up before, at least you have an idea of where to look for data or how to revive data or try to figure out where. It is, but it's. It's another waste of time. You know you're you're going to a location where either something should be or is not exactly how, as it should be. I mean, a lot of times this comes down to to call a meeting, you e-mail someone, you call someone or you walk, you walk to different building and you start asking questions. And anytime you can do this one, it's it's inefficiencies. So that's kinda how it starts, where you talk to. Someone that's been there for 20 years, you know, there's a lot of tribal knowledge that goes on of just how do we solve this problem? What could have gone wrong with this piece of machinery? Or how it was uploaded to the system. It's it's very time consuming.

Vivek Vishwakarma: (20:55) Do you think is there any tool out there to capture the tribal knowledge?

Stephen Drake: (21:00) I've not seen a good one. I I know it's a larger concern for some of these bigger companies. It's weird. I'll see it as a big concern. It's a very large companies and they're very small companies. So in the very small companies, you typically have one such you like single point failure. If you want to call it that of just one guy that's been doing this one thing for 20 or 30 years. So if.

Vivek Vishwakarma: (21:16) Mm hmm.

Stephen Drake: (21:19) You can kind of get around that. It's helpful, but with a smaller company you don't have all the resources to be able to to get all that knowledge. They always say they're gonna do it. You know, as someone realizes they're going to retire in one year. The idea is, let's get you know, let's download what we can from the seed in person. But it just doesn't really happen. Because again, to the first point, you have to start making progress and it's about doing something quickly and making progress so. It it typically goes away, but it's it's a lot of just communication with people of just where do we start.

Vivek Vishwakarma: (21:52) Right, right, right. I mean as we are getting closer, I do want to to touch upon one more thing is. I feel like we're living in a very interesting time, right? AI is just not a buzzword. It's gonna be here. It's gonna stay here, but the big question is how do CIOs and you know the leadership team see it getting integrated with their sops, right? So what are some of the emerging trends and technologies that you are more personally excited? And think is gonna transform the space, be it from the data management data loss or you know troubleshooting in general.

Stephen Drake: (22:43) Yeah, I I agree. I think from a management point of view. These the smart companies realize that this is going to be a large gamechanger and it it's going to take some time. You know, I don't think anyone's under the impression that's going to revolutionize everything for for most industries in six months or one year. But they need. To invest, they need to understand it. I feel like technology is best used when you use it appropriately. Every technology can do something amazing, and it does a whole lot of things. Not so good. You know you wouldn't use a curling iron to toast bread. You know, something like that. So just use it where it's appropriate. And same thing with same thing with AI. It always had a great potential. You know, some people used it as a buzz word, but it actually did have really good practical uses. And so using it to its best abilities is helpful. I think the thing I'm most excited about is to actually start seeing this getting implemented. You know the best way to get something widely adopted is to make it absolutely easy to say yes to, to be able to get integrated. You know, when you you go back to the call it buzzword for five or ten years ago, you know IoT or Iiot we. Were told what it was, but getting it adopted and integrated for for the bulk of people, it requires a lot of training, a lot of data, a lot of know how.

Vivek Vishwakarma: (24:09) OK.

Stephen Drake: (24:10) The promises of what it can deliver were massive. But actually just doing it, getting it accomplished. It took some efforts. You know, it took more than one person more than one day, and so the ability with AI is it kind of remove some of those steps. It's kind of like these graphical based coatings, you know for computer languages.

Vivek Vishwakarma: (24:27) Alright, alright.

Stephen Drake: (24:27) You know, anything you can do to to get the same end result, but to do it faster not and not just to do it faster, but also to kind of minimize the the hesitation of the IT departments. You know the the overlords of our computer systems. And and there's concerns with what AI can do. So as long as you have limitations on what is capable of either reading or writing or executing, or or even just segments of where it can live.

Vivek Vishwakarma: (24:51) Right, correct. Correct.

Stephen Drake: (24:54) But just just seeing it being used practically, you know, not just seeing it on the Internet for apps, which is helpful and fun, but seeing an industry where it can actually make a difference in manufacturing. There, there's a lot of things that can be done. Infinitely more efficient small pieces here and there. We need to do it across a wide variety. I think it can really make. A big difference, and you know it's it. Maybe it's 1% a year, but five or ten years now it's gonna be more than that one. And you can actually see a big, big difference.

Vivek Vishwakarma: (25:26) No, this is great. This is great. I think this has been a great, you know, conversation that we had. I would like to thank our listeners and thank you, Stephen for you know shedding light on the data loss and how how it impacts the industrial troubleshooting and how it's gonna look like for the next 4-5 years. So yeah, I mean, I'll catch up with you again in our next podcast, hopefully with couple of more speakers. Tell them. Let's wrap this up.

Stephen Drake: (25:59) All right. Thank you.