The $50K use case matters – Zev Arnold on Scaling, Context and AI

David (00:00) Welcome, you’re listening to the IT/OT Insider podcast. I’m your host David, subscribe to our blog or a podcast to learn more about our work or go to itot.academy to discover our training. And today I’m joined by Zev Arnold. Zev is a principal director at Accenture with a focus, but also a passion, I have to say Zev, on a digital manufacturing operations. So welcome to the podcast. Zev Arnold (00:25) Thank you so much, David. It is a pleasure to be here. ⁓ I’ll start by just reintroducing myself. So my name is Zev Arnold. I’m a leader within Accenture’s Industry X business, which is the business within Accenture that helps our clients reimagine the things they make and the ways they make them. I’ve spent my career working mostly across process manufacturing industries, oil and gas and utilities and energy and mining and life sciences, helping engineers and operators use data to improve the ways they work. ⁓ so that’s me in a little bit of my background. I’m, thrilled to be on this podcast. I’ve listened to all your recordings. I think you guys are doing amazing work. I understand you might have a book coming out in the not too distant future, which I’m very excited to read. ⁓ if I’m allowed to say that and, you’re to pull that out. ⁓ and I’m also, I am also really excited to see, ⁓ what comes of the, your talk in Las Vegas, which you did announce publicly on LinkedIn very recently. David (01:10) No you’re not, but it’s okay. ⁓ Zev Arnold (01:24) Anyway, I’m thrilled to be here. Thank you, David. David (01:24) Yeah, cool stuff, cool stuff. I’m not going to cut it out. We’ll do another announcement ⁓ some time later. But our paths have virtually crossed already a couple of times. the thing which stood out for me, ⁓ Zev, is, as I said, your interest in, I would say, especially ⁓ process manufacturing, but also discrete. could we like, could we like start, maybe for the listeners of the podcast, could you like start? Zev Arnold (01:30) I’m not. David (01:52) by explaining a bit those industries, bit, say, your stories, your takes on what it is to work in those super important industries, they’re all around us. Zev Arnold (02:05) Sure. Yeah, thanks, David. Let me start with the lay of the land as I see it. So the lay of the land according to Zev. ⁓ When I use the term manufacturing, I typically use like big tent manufacturing, including almost any industry with heavy assets that runs ⁓ a factory or plant monitors equipment. ⁓ These are the industries I’ve worked in and they’re interesting because they have a large amount of what I would call operational technology data. which is distinct from IOT data, right? My Fitbit or my Nest is an IOT device. This is different. This is sensors and plants, and there’s a lot to distinguish how we work with and use the data from sensors and plants. So we think about manufacturing industries, all industries with plants and equipment, sensors, operators and engineers. What I came to realize over the course of my career is that there’s a… important distinction between discrete and process manufacturing. In discrete manufacturing, we have production lines. They are making a manufactured product. ⁓ Typically, the manufacturing lines are very similar from plant to plant. They have the similar process that they’re running. ⁓ These industries tend to use MES systems to help them manage ⁓ recipes and orders. And it’s the kind of industries that Walker Reynolds talks about a lot in industry 4.0. I’m a big fan of too. So discrete manufacturing industries, they make products for people in general. Process manufacturing is different because you’re not making a particular product or a discrete good. You are running a continuous process, upstream oil and gas. You poke a hole in the ground, the oil comes out and you just keep the oil running as fast as you can. It’s a little more complicated than that, but that’s a continuous process. David (03:32) Mm-hmm. you Zev Arnold (03:58) Transmission distribution and electricity is a continuous process. have electricity coming onto the grid and you are distributing out across the grid in a way that balances the grid and it meets your voltage requirements and minimizes lost power. So when you have a process manufacturing industry like the industries I’ve worked in, your focus is on asset reliability and process efficiency. and sustainable operations. And while all three of those things are important in discrete manufacturing, they have a slightly different meaning in process manufacturing. So I’ve come to recognize over the course of my career that as we talk about industrial IoT transformation, which is how I describe this era that we’re moving into, it has different implications for process manufacturing and discrete manufacturing from the technologies we use to the use cases that we pursue to the value we deliver. David (04:38) Mm-hmm. Zev Arnold (04:52) to the adoption by the business, to what the business transformation actually means. And so it’s become a useful distinction ⁓ for me to understand how I put my experiences in the context of the broader experiences that manufacturing, Big Ten manufacturing, describes. David (04:57) Yeah. And if you then, if you reflect on those experiences, like are there similarities, are there, I would say similarities in failure patterns, are there similarities in success patterns, or ⁓ should we really treat those sectors differently? Zev Arnold (05:34) I think so. So I. I can’t claim a lot of expertise in discrete manufacturing. I just haven’t had as much work experience there as process manufacturing, so I can talk with certainty about process manufacturing and what I see in process manufacturing is that we have significant challenges with data contextualization. We don’t have MES systems as a as a starting point for data contextualization or using our operational technology. And the. ⁓ the use cases tend to be broad and shallow. And so what I mean to that is that ⁓ in discrete manufacturing, at least where I’ve seen success with industrial IT transformation, discrete manufacturing, the use cases have been few and deep. We are gonna optimize ⁓ the packaging of candy for this particular kind of candy on this particular kind of manufacturing line. And then we’re going to roll that out across every manufacturing line for that particular piece of candy across the entire world. It’s going to be very similar in the rollout and it’s a high value use case. In continuous manufacturing, we tend to end up with a situation where, there are those deep use cases. ⁓ Like for instance, Shell has a presentation they did with Accenture several years ago about blowout preventive reliability. And they built a application. can Google this and find it online. They built an application to manage the reliability of their deep water subsea blowout preventers. And that was something that they could scale across their entire deep water fleet. And so those use cases do exist where you can build once and deploy. But I’ve come to believe that the bulk of the value is not from these few deep use cases, but rather from shallow broad use cases. So I’ll give you a different example. There’s a hydroelectric. ⁓ David (07:26) Yeah. Zev Arnold (07:30) power generation company I worked with where they had these sump tanks that would periodically overflow and cause a pollution incident downstream, which would cause fines of up to $50,000 a year. Saving $50,000 a year and avoiding the impact on the environment is well worth doing. This is a valuable thing. I can’t spin up a $1 million Accenture digital transformation team to go save $50,000. Right? That doesn’t make any sense in the world. So. But what we could realize is that this is just one of maybe hundreds of examples of opportunities to use data to make better decisions across that enterprise that altogether add up to tens or hundreds of millions of dollars in annual savings. How do we go get that? And I think that’s been one of the unique struggles in process manufacturing that I’m not convinced has a synonym in discrete manufacturing. I’ll say the jury’s out. I have opinions. David (08:14) Yeah. Zev Arnold (08:29) But I’m to keep those opinions to myself until I know more about what I’m talking about. I’m interested if you get any reaction from your followers on the podcast, I’d love to explore that idea more. But to sum it up, are real differences between process and discrete manufacturing. And I think we need to be a little bit more careful as a thought leadership community when we talk about industrial IoT transformation in thinking about how the things we recommend from technologies to change management processes to… David (08:30) You Zev Arnold (08:58) approaches to transformation, how those might be different between those two different sectors. David (09:04) But and that’s an interesting example. it’s also you like briefly mentioned, for example, the fact that MES systems typically don’t exist ⁓ in a process environments, is is which is really it’s absolutely an interesting fact. ⁓ I made the I made the other change. So I have this most most listeners know that I have my 12 years experience in the chemical industry. ⁓ And then, you know, we we use the term MES. So I thought I knew the term MES, but then I changed over to food and bath companies and I went like, yeah, that’s total. Zev Arnold (09:40) It’s a different world than food and bed. It’s like how would you describe? I’m interested in the difference between the problems than MES system solves in food and bed versus what it solves in chemical manufacturing. David (09:52) It’s a really interesting one. think for food and beverage, it typically starts with traceability. ⁓ So I think that’s the like the, yes, you can do much, much, more with that, but it typically starts from some kind of a traceability needs. ⁓ and then obviously to optimize your lines and then your operations, et cetera, et cetera. But it comes from traceability where in a process context, as you said, you opened the well or in the case of chemicals, you start up your reactor or your distillation column or your train or whatever. And at that point in time, Zev Arnold (10:28) I am never gonna have to recall a hydrocarbon because of a quality issue, right? That’s just not something that would happen in an oil and gas company. Whereas Food & Bev of course needs that traceability. David (10:38) Yeah. No, what you would do of course is yes, you would do your ⁓ maybe online or offline ⁓ checks of your, I don’t know, the purity of your hypercarbon or whatever. And you might well decide to start blending some pure product with some less pure product or whatever the thing might be. But you would actually never ⁓ indeed either do this recall or. thinking in terms of batches or runs. I think the only thing I’ve seen which resembles an MES system, the closest is for example, in the case of let’s take a cracker where your ovens, they have runs and somehow you wanna visualize ⁓ the, I would say the runs of your ovens because at a certain point in time, you wanna for example, decode them or whatever. And then I would say that’s kind of. a badge, I don’t know. The closest I could. Zev Arnold (11:38) So there’s. Yeah, I mean, it’s an interesting spectrum from any S to batch monitoring to fully continuous where a batch doesn’t even necessarily make sense. So in ⁓ electric distribution, for instance, I don’t know exactly what a batch would entail. You do have operational events which are of interest. So, so for instance, one of the problems that I’ve worked with a client on in ⁓ in distribution is. David (11:58) Mmm, yeah. Zev Arnold (12:05) automatic reclosers. So these reclosers are the reason that when you lose power, it immediately goes back on again, because they reconfigure the grid automatically to get power back to your house and isolate the short wherever it is. So instead of 1000 people losing power, just one guy loses power. That’s unfortunate for that one guy, but there are the 9999 of the rest of us, you get to continue to keep the food in our refrigerator are very happy about that. I live in New Orleans, so we deal with the power outages. very much in my mind, it’s power outage season right now actually, but in any case. ⁓ So for them, an event is, show me all the time that this recloser was from when it engaged to when it disengaged. And that is an event of interest, an operational event, which is very different in nature from a batch, right? Superficially, they’re somewhat similar. They have a start time and they have an end time. The use cases around them are entirely different. What you’re trying to do with the information and David (12:54) them. Zev Arnold (13:00) the metadata around it and everything else. So that’s a great example of. David (13:05) It’s an interesting exercise to just talk about or think about these differences. to come back to ⁓ what you mentioned, you mentioned the 150k example and then you might find another potential 20k example and maybe after a while you also find a bigger one. to able to do that type of things. in our articles and our podcasts, we typically talk about scaling. So in order to get there, you need a way to be able to reach that scale or that ease of use to not have ⁓ to invest that million dollar in a big project to get a 50K outcome. No, you want to have some kind of a foundational layer. ⁓ So let’s talk a bit more about that. Zev Arnold (13:57) That’s right. That’s right. So you need a common platform. one of the, there’s a lot of challenges that I’ve seen over the years as we’ve attempted and succeeded less than we may be planned to in industrial IoT transformation. I’ve seen so many companies over the years, right, take on big initiatives and they don’t deliver the value that they could. And so where are the challenges that we’re facing? And one of the big challenges is exactly that, a common platform. So there was, there was one project I was aware of where we delivered a few use cases ⁓ to this company on a, on, on a platform that we built for industrial IOT. It was very successful. I’m not going to go into the specifics. The point is, is that the challenge that project faced over time is that adding each new use case was every bit as hard as building the first use case. because identifying which platform components to reuse was very challenging in the moment. So I would build a use case and I would bring it out and I would build it for that particular use case. All of my data collection and modeling and transformation and calculations and analytics and the whole deal, visualizations, right? I’d build that out and then I go to build the next one and these components would not be. David (15:13) Yep. Zev Arnold (15:19) exactly micro serviced in the way I exactly needed to go build the new use case. And so you end up rebuilding a lot. The industry, I mean, if we look across the technology in the industry, I see all the software vendors attempting to solve this problem, trying to figure out what are those reusable components that ⁓ form a base of a platform. In process manufacturing, one of the most successful vendors in this category has been AVEVA. with the AVEVA Pi historian, formerly OSIsoft. Is it okay to name specific vendors? Can we name names here? David (15:52) Sure, And they were already on our podcast a while ago, so why not? Zev Arnold (15:57) They were okay. Very good. So it’s their own fault. ⁓ what, what AVEVA done really well is go beyond what are the common platform components for a technology team to use to go deliver a use case. And instead what they’ve asked is what are the common components that an end user would want to pull together on their own? So the, approach I’ve taken with many of my clients. David (16:00) You Zev Arnold (16:22) an approach that I think needs to be more broadly accepted in the industry is to enable self-service directly to engineers and operators. When we talk about that $50,000 use case, it doesn’t make any sense to have an IT team dedicated to going and building that solution. That’s got to be the engineer who understands what the problem is, who understands and knows how to get the data, how to model the data in the right way, how to build the analytic, build the event that gets triggered. David (16:29) Absolutely. Zev Arnold (16:50) write the notification around it, track it over time, all the components in order to do something about this and get out ahead of the problem before you have an environmental incident, that all has to be at the disposal of the engineer who truly understands the problem. And if we take a step back abstractly, what this really is is a stronger collaboration between people and AI. That’s what we’re talking about. It’s maybe not, you know, here we’re talking about maybe I’m monitoring the level of my sump tank to make sure it doesn’t go over a threshold? Well, that’s not generative AI. It’s not even machine learning, but it is an expert system. It’s the most basic form of AI. How can we put expert systems in the hands, know, tools to build expert systems in the hands of engineers and operators so they can collaborate with those more effectively? We haven’t even answered that question yet, much less how do I put machine learning in the hands of engineers and operators to build their own tools, much less how do I put gen AI in their hands to build their own agents? I think we might be getting there, and it’s a very exciting time with Gen.ai. I don’t know if you’ve heard of these large language models. David (17:53) course it’s like ChatGPT is my is my best my best new buddy Zev Arnold (18:00) ⁓ My wife ⁓ has a whole list of chat GPT agents she’s built that help her with different problems in her life. Like this is my agent that helps me with work. This is my agent that helps me with scheduling the kids and packing for trips. right, like that approach to work, I think we’re getting very closely to where that becomes ⁓ quite plausible. But the overall idea is that the way to go after and achieve all of this value is to build that. David (18:07) huh. See you. Zev Arnold (18:28) relationship between human and AI so that it is more natural and more self-service and more ⁓ spontaneous. And we get large development teams out of the loop of building these solutions and more focused on supporting the platforms. And I think that has to be our direction, at least in process manufacturing. again, mean, coming back to my core conceit here of the differences between process and discrete. David (18:30) you Zev Arnold (18:57) Is this as relevant in discrete manufacturing? And I don’t know. And I’d love to hear. ⁓ But certainly within process manufacturing, within oil and gas and mining and utilities and perhaps even to a certain extent chemicals and life sciences and in our batch process manufacturing or like hybrids, there’s tremendous need for that, I think. David (19:19) I’m confident that there is need across the board. ⁓ But indeed there are some sometimes subtle but also sometimes big nuances or big differences. the interesting, I just wanted to because you are now talking about this people or human AI interaction thing. So this is really interesting because yes, I also see lots of things happening right now. the number of startups currently active. Zev Arnold (19:23) But David (19:48) even just, and not even talking about the IT domain, but just the industrial IT domain. So startups are active in the industrial domain and even scale-ups, that’s huge. I think they, I actually had that conversation yesterday with somebody where I said like, it’s actually becoming super difficult for all these startups to differentiate themselves between each other. So that’s actually a big problem now. ⁓ For those who haven’t subscribed to the podcast yet, me and Willem, are actually preparing an entire series on industrial AI, which is planned for, ⁓ I would say starting in September. starting in September until more or less the end of the year, we will be ⁓ publishing a lot of industrial AI content. So ⁓ we’ll also come back to that point. So that’s just a small, yeah, yeah, we as well. And we really wanna demystify some things and… ⁓ Zev Arnold (20:39) very excited about that. I will stay tuned. David (20:46) and talk a bit more about why. I’m not really about the technology side of things, but really about the why question. What are we trying to answer? But ⁓ one thing I think every data project has in common ⁓ is data modeling, data governance. ⁓ There are certainly differences between industries, but they all face the same problem. The problem of different lines, differences in sensors. know, even I remember one company who really this multi-year standardization project only to get to a point where the CEO announced a big merger and then they could actually start over again. They could start over again because all of a sudden they saw a totally different set of devices, et cetera, et cetera, coming into their network. ⁓ But what have you seen around data modeling data governance? What works, what doesn’t work? Zev Arnold (21:49) Sure, this is a great setup. I’m going to tell Mike, this is my favorite anecdote. This is my origin story. Like how Zev became fascinated with industrial IoT. I started my career as a AVEVA PI system, like a support analyst. I would work with IT organizations to do, you know, system upgrades, connect new devices. David (21:59) Okay cool, tell me. Zev Arnold (22:15) make sure the system was running and stable, provision users with access, right? All the sort of core IT lifecycle stuff around supporting AVEVA PI which is for those of you not familiar with it, it’s a commonly used historian system or platform in process manufacturing in particular, discrete manufacturing as well. ⁓ I convinced myself that my focus for my career would be excellence in operational support for AVEVA PI and I did not need to know what the end users did with the data. It just wasn’t my concern. And so then one day this guy walks into my office. He was a compressor engineer and great guy. ⁓ And he asked me, he was like, look, ⁓ I’m trying to improve the way we do maintenance on our compressors. We’ve just been running our maintenance on our compressors from a schedule for years. We just do it every X months, whether they need it or not, and we don’t know. we would like to start looking at compressor starts because the biggest wear on a compressor is when you start it up. So if we can figure out how often we’ve started it in the last period of time, we would know whether it really needs maintenance, which is very expensive to do, especially on offshore platforms or not. ⁓ And then the challenge, and this is how comes back to data governance. Some of those compressors had tags that indicated whether they were running or not. And so relatively easy to set up a… ⁓ David (23:04) Yeah. Zev Arnold (23:32) analytic that would look at that tag for each compressor properly modeled and say, it running or is it not? And then count that up over time. Some of the compressors didn’t have that tag, but they did have a rotation speed. And so I can convert the rotation speed into essentially a virtual counter for starts and stops. And so this, this fascinated me. And this was the turning point in my career when I realized that what I really want to do is help drive value in industry, not just support the systems. I think those two things are. They have to be paired. have to be done together. You have to have operational excellence focused on user value. The other thing I discovered from that is that the structure of the data, the contextualization of that data to serve a use case for an end user is only truly understood by the end user who needs that use case. If we think about this particular engineer, he needed a way to model the data so that he had compressors with a cycle count. David (24:00) Yeah. Yeah. Zev Arnold (24:27) underneath them as an attribute. And that cycle count had to be derived from other source tags, depending on, on, on where those tags were in the system. If all I had done is come at it from a more naive approach of I’m going to model this data by following this like Purdue model of organization and just say, here’s all the tags for this compressor and never really think about that and use case. I’m not giving him the tools he needs to do his job. And so the core to contextualization I’ve come to believe. David (24:52) Mm-hmm. Zev Arnold (24:57) is it needs to be driven by a specific use case. It needs to be in service of a use case, and it needs to be owned by engineers and operators to create that contextualization. And the way to make anybody in the business own something is to make them care about it. Because if you just give them a task on their performance evaluation, please govern this hierarchy, that will never get done. You need to give them steak. Right. Why are you organizing this hierarchy? Because I need to deliver this use case, right? Give that hierarchy to that compressor engineer and say, this is your hierarchy. Now, if you want to be able to manage your compressor maintenance, you need to, when a new compressor is entered into service, added into this data model. And then I need to have tools so that when he has it into the data model, he can do that in a way that’s intuitive to him and doesn’t require writing a bunch of Python. And that once he’s added to the data model, the analytics just start to work and he doesn’t need to rebuild any. I mean, got to make self-service easy. David (25:38) Mm-hmm. Zev Arnold (25:54) And then come back to contextualization. Contextualization needs to be self-service driven by the business, governed centrally. David (26:02) It’s an interesting one and especially the governed centrally, it’s a very, it’s a, maybe that’s the million dollar question. how do you embed that in an operational organization? And should there be some kind of an IT data governance and also an OT data governance or can they converge question mark? Will they converge? Do they need to converge? I’m personally still struggling to answer that. ⁓ I don’t know whether you have some thoughts, but for me it’s like… Zev Arnold (26:40) I, I absolutely have answers. have multiple answers and they’re all contradictory. ⁓ so let’s, let’s just pick one. as we think about industrial IOT transformation, we haven’t talked a lot about the value of IOT transformation. I kind of wanted to talk about that a little bit. We can come back to it, but let’s assume for the purpose of our conversation that there is value in making better decisions with our data. Let’s just take that as an assumption for now and we can maybe dig into that a future podcast. David (26:44) Thank you. Yeah. Zev Arnold (27:08) ⁓ if that is true, and if we want to go capture that, we need to transform the way that people work with people and people work with AI and AI works with AI within a plant. The holistically part of that transformation in changing the way that people work with AI is around this data governance question. It’s, it’s part of changing our business functions. need business functions within the business, engineers and operators who, who David (27:35) Mm-hmm. Zev Arnold (27:38) Accepted as their job. My job is to model this data and build analytics. I’m an engineer, but my job is to ⁓ train and maintain AI agents. And to borrow the modern parlance of our times more simply, right? AI agents could be expert systems. It could be as simple as I want to monitor my sump tank level so that it doesn’t overflow against the threshold. A very simple AI. But still it’s got to be somebody’s job to train and maintain that AI expert system, no matter how simple it is, or no matter how complex it is. That’s going to be your job. And that’s when I think about industrial IOT transformation, I think about it in those terms of, of changing all of those relationships, not just people to people, not just people to AI, also AI to AI and what that all means for what the future of our business functions mean in a plan. David (28:15) yeah. Zev Arnold (28:35) think there’s a lot of room to explore and grow and mature our understanding of the best way to run a power plant or an electric grid or a deep water production facility or anything in the process manufacturing space and arguably across the discrete manufacturing as well. David (28:48) Mm-hmm. But I like that people, people, AI, AI, AI, ⁓ yeah, thought or even see it’s even already it’s a framework. I might borrow that framework for some future articles. Zev Arnold (29:06) So it’s fine. It’s well used. It’s not mine. So actually a lot of my thinking on this has been inspired by Paul Daugherty who was the former chief technology officer of Accenture. He wrote a book called Human + Machine which transformed my understanding of what the AI revolution or the second machine age might mean to industry and to people and to jobs and to enterprises. ⁓ was the… The thinking from that book formed actually the basis of my most recent presentation at AVEVA World. I frequently attend the AVEVA World conferences. I use that as an opportunity. You and I were discussing this earlier to collect my thoughts for the year and present them out. in my last most recent presentation, which unfortunately isn’t on video, so you’ll just have to take my word for it, I focus in on his model of what a human and AI collaboration looks like. David (29:47) Yes. Zev Arnold (30:00) And so I’d encourage people to go read the book. I thought it was fantastic. And it certainly formed the basis of my thinking on what might those new job functions look like, especially in plant operations and engineering. David (30:13) super interesting. I’m gonna look up the book and I’ll make sure to add a link to it in the show notes as well so if people are interested ⁓ they can easily find it. ⁓ As a last short question, maybe even as some kind of a outlook maybe for another podcast, you said let’s assume there is value. Can you like… Should we assume there is value? ⁓ Are we looking for the right things or? Zev Arnold (30:47) On my worst days, when everything has gone wrong for me, I wonder to myself, is this data really valuable? Do we really care about this? Right? And over the course of last 15 years, I’ve convinced myself absolutely we do. ⁓ Here’s a concrete example. This is the concrete example I used in my AVEVA World presentation back in November, ⁓ April, I don’t know, whenever it was. The US power generation industry. David (30:50) You Zev Arnold (31:16) Right. Power plants, right. Nuclear, coal, hydroelectric, solar, wind. They have a rating that’s publicly tracked and a monitor called E4, equivalent forced outage rate. It tracks the amount of time that the power plant has been unavailable to generate power due to some reason that we wish hadn’t happened, whatever it is. Right. It’s power we could have generated, but something went wrong. So we couldn’t. It’s on average about 7.5%. opportunity to reduce 7.5 % to 0%, right? E40 is, ⁓ I think it’s a hundred billion plus opportunity just in additional electricity revenue to power generators. Not counting the additional opportunity to transmission distribution, not counting the reduced reduction in electricity costs to consumers and industries. know, all of that, that David (32:01) Yeah. Zev Arnold (32:14) That headline number of $100 billion is just one aspect of the value, which then balloons through society and allows us to have better data centers and allows us to have a climate friendly mix of our energy supply. mean, just the domino effect of that just ripples through society and turns that $100 billion of value into a trillion dollars of value, or who even knows how much. And that’s just one example in one industry. David (32:29) Mm-hmm. Zev Arnold (32:43) It’s things like that that have convinced me there’s real value to be had. Like E4.0 is a tough goal, but we know that we can do better. We as an industry know we can do better than what we’re doing today. We know that we have outages that we can avoid if we can make better decisions with data. And so that’s what I continue to plan doing for the next, let’s say 20 years of my career. We’ll see how long it goes. is to continue to work with industry to make those improvements, to capture that value, and in the end, deliver that value out to society so we can tackle some of the issues that I think is what we’re all here together trying to tap on. Like, how do we solve our energy prices? How do we solve climate change? How do we do this in a way that provides middle-class jobs, right, and doesn’t gut out the middle of our society? I mean, there’s so many societal impacts that we’re not going to get into in three minutes, but… ⁓ But that’s what gets me excited. That’s why I’m so excited about this space and why I’ve dedicated my career to it. David (33:40) This is a perfect, it’s also a super, super, super interesting example, but it’s also a perfect way to end this episode. ⁓ Let’s plan for another one in maybe a couple of months or so. ⁓ Love to pick your brain a bit more. ⁓ So Zev, yeah, thank you for joining me. Zev Arnold (34:01) Thank you so much for this opportunity, David. It’s been a pleasure. David (34:04) And of course also thank you to our listeners for tuning in. If you enjoyed the conversation don’t forget to subscribe at itotinsider.com and discover our live online training at itot.academy. And see you next time for more insights on bridging IT and OT and until then take care, bye bye.