From Platforms to Outcomes: Vatsal Shah on What Industrial AI Actually Needs

David (00:01) Hello and welcome to the ITOT Insider Podcast. New episodes. Very, very interesting guest. ⁓ Vatsal, welcome to the show. Vatsal Shah (00:13) Absolutely. Thank you for having me David. Really appreciate it. David (00:15) So this is actually the the first time for those who are watching, the first time I do a recording in my new office with a new background. It's not a virtual, it's a physical one. I can touch a thing behind the pinball machine behind me. ⁓ still a bit of work in progress, so things might change in the months to come. But I'm I'm super excited to have this new this new venue and having with with me here Vatsal Shah, the CEO, ⁓ founder of ⁓ of Litmus. Vatsal Shah (00:42) Yeah, no. ⁓ Thank you. And just for everyone, we nerded out for that pinball machine for 10 minutes before joining you. David (00:51) Yes, yes. You know, the thing is full of electronics. And ⁓ and so this one is ⁓ a late 80s, and I always I always made the same stupid joke. Is this is still electronics? Like you can understand visually. It's and it's I I I love it so much, but anyways, that's another it's it's another episode. You know Vatsal Shah (01:07) Yep. Yeah. Yeah. That's a whole other podcast. David (01:15) You know what we're gonna do? ⁓ in this one, we're gonna talk about ⁓ the other side of the spectrum. so not visual electronics, but but more complex artificial intelligence and what is happening. ⁓ but before we do that, Vatsal, your introduction please. A bit about yourself, a bit about Litmus. Vatsal Shah (01:34) Yeah, so first of all, thank you. Thank you for having me. I'm Vatsal Shah I'm co-founder and CEO at Litmus. Engineer at heart. Still do engineering and ⁓ still love what engineering and software can bring to manufacturing side. Started my career on the plant floor. My first job was on the power plant and then food and beverage, oil and gas and everything else. Founded Litmus close to 10 years ago. We are ⁓ a bunch of extremely passionate guys at Litmus trying to solve ⁓ something that was very difficult to solve or unsolvable for a startup. we have been like, I'm quite lucky to have good co-founders, good engineers and good team by our side. And we are revolutionizing the whole industry as a whole. that's David (02:22) It's it's not a joke. It's not a joke because I think every time I talk to anybody in your organization, it's like it's like talking to technical people, right? It's it's I I love that. I love that and also love it that you are ⁓ in our in our in our show. Now ⁓ people should know Litmus. ⁓ we wrote about ⁓ you guys a couple of times ⁓ in in our blog, but maybe the elevator pitch, ⁓ what are you solving? Vatsal Shah (02:31) Yeah, yeah. So Litmus is the foundational data platform for the industrial world. So we help large, extra large, any different type of manufacturing companies standardize and harmonize data on the plant floor. Today, you have existing pre-existing applications from SCADA system, MES system. Tomorrow, agentic world and AI and cloud and everything else is popping up. How do you have complete collection, connectivity, and analytics of the data on the plant floor? That's what Litmus does. So customers are trusting Litmus when they want to scale from a success on a line number one to 200 sites or 2,000 sites when they want to scale it up. Litmus is where we are very successful there. David (03:32) And ⁓ I would say to to our audience, a little over a year ago ago, we did a podcast with John, John Younes where we I would say stepped into the data ops ⁓ storyline ⁓ and the data platform storyline. Maybe we touch upon it a little bit more, but in this one in this one, I would like to look forward. I would like to take a look to it is It's actually an in it's interesting because obviously everybody is talking about AI. The fun story here, Vatsal, is so I was at Hannover two years ago, I was at Hannover ⁓ this year, and like two editions ago, I had the impression that I wouldn't say literally every boot, but almost every boot had something on. We are the industrial AI or the AI platform. We and we enable AI, buy our stuff, and you have AI. For some reason, I now see that conversation changing a bit too. Let's first start with a data foundation. And let's then start layering on top of that. like like w how do you see the the the current storyline? Vatsal Shah (04:39) Yeah. Okay, so this is the first time I'm talking about this publicly ever. let me share you our thought process. We thought about it a lot, lot. ⁓ Like what we learned was ⁓ there are 10,000 plus sites there trusting Litmus on the Plant Floor right now. It's again, we are one of the leading vendor in the market. We are the platform for industrial data platform out there. So we are out there, we are deployed. ⁓ What we started understanding, David (04:59) Okay. Mm-hmm. Vatsal Shah (05:19) from our customers is how they are consuming data out of Litmus platform. That was our step one. In 2023, 2024 timeframe, we asked our customers, we looked at the platform that they were integrating with us with, the type of application, type of KPIs, type of analytics that they were building. And we have a humongous dictionary of everything that is possible with the data platform. That's literally step zero that we took. And that was a couple of years ago. So we know what's going on on the plant floor, in the cloud, in the AI world altogether. Again, that learning helps. But then we, rather than creating one more platform, we thought about what do we want to do? Like, how do we want to help? How do we want to help these customers get where they want to go faster? That was our question. And we wanted to work back from outcomes to the technology rather than technology and then deliver outcomes. So in 2025, crafted a perfect strategy, let's say perfect strategy on paper at least, strategy for Litmus. If you look at manufacturing software market, approachable, accessible, it's what, like $15 billion market? You look at the data and software itself. David (06:22) Mm-hmm. Mm-hmm. Vatsal Shah (06:47) It's small. How much is the manufacturing labor market? A trillion dollars plus. It's huge. It's a hundred times the manufacturing software market. Not five times, not ten times. It's a hundred times. Why all of us, we are fighting for a small piece of the puzzle became a question, which is, yeah, Litmus is growing. One of the fastest growing in the market, in the market right now. We are growing. David (06:52) Я ча. Vatsal Shah (07:16) To double down our revenue is not even a question. We can keep on doing the same things and we are going to have 100 % year-over-year growth. But our idea was how can we bring significantly more value from the technology? And no, we did not know the answer to that one. So we went into soul searching mode that how do we deliver the outcome to the CFO, to the CEO, to the COO of a large manufacturing company and what matters for them? And while we were thinking, agentic and AI agentic systems, they were evolving at the same time. So now going back to that trillion dollar labor market or two trillion dollar labor market, it's ⁓ Can you build a system that augments the existing human being on the plant floor and take 20 % of their job and help them do something more intelligent? Like just taking 10 % of their job. And can this AI system on the salary of that 10 % salary of that human being? If answer is yes, you are not selling technology at that point. You are selling outcomes. Like what humans are driving, agentic AI systems would be driving the exact same thing. So we created this narrative, and we created a bunch of this agents. We never put it in production. But we were saying that there is a quality advisor. There is an energy advisor. There is David (08:22) Mm-hmm. Vatsal Shah (08:41) There's like agents that we have right now, but they were not successful on the plant floor. Then we figured out, OK, for us to make that open-claw, Nemo-claw successful, we require an inferencing system. So let's launch that technology. We require a data catalog for metadata. Let's launch that technology. We require agent-to-agent communication or MCP communication or augmentation of SLM and LLM. So we kept on building technology pieces. to make sure that those Spotify agents were successful. And that's what we did. So rather than building the technology, we figured out what outcome we want to drive, drive it and then build technology to fulfill that outcome. So we work backwards from user experience. David (09:20) Yeah. That's that's that's that's a very interesting take and it it also kind of you know ⁓ me and Willem ⁓ often joke like we don't need another chatbot. and and and and the reason is the reason is exactly a as as you described it, like what what is the value of me asking a random question ⁓ about my plan to the chatbot that that's that there is the there is no point in doing so. Vatsal Shah (09:39) No, you're not. Yeah? David (09:58) especially because ⁓ and and and then we come to the ⁓ we wrote about that a while ago to the two main yeah problems or reasons or or things we need to solve industrial AI is first of all the the scalable data foundation, the ⁓ something and we talked about that so many times, Fatal, what what we then later called our our our industrial data capability map ⁓ in in its in its full extension. And then secondly, ⁓ If you are then applying AI models to manufacturing problems, I'm not talking about applying them to labor ⁓ augmentation, but to manufacturing problems, then you also need to understand the physical reality of the of the manufacturing plant you are talking to. Vatsal Shah (10:42) Yeah, exactly. this is the whole point that everybody, Frontier Labs are investing $100 billion in creating AI and agent ops and everything else. Google is investing, Anthropic is investing, Microsoft is investing. All of these companies, are investing so aggressively to make sure that agents are capable. But it's so simple. For those agents to work, for those AI services to work, you need data, you need meaning of that data, which is context. And then you need ⁓ everything that goes on, actionability, improvement, and everything else that you have to build. You create this gumbo, and you make it available to AI agent, it might actually do something meaningful. So you have to figure out, you have to work backwards from the outcome that you want to drive, and whatever technology is required to fulfill that. And that was one of the reasons why we wanted to launch the Industrial Data Catalog in the first place, to drive that out. And in your book, I think you talk about, ⁓ like in one of the chapters as well, the platformized approach. ⁓ Today's applications, are purpose-built. SCADA is purpose-built. MES is purpose-built. ⁓ It's a human-built. Again, not so long ago, we used to write the code by ourselves with our own hands. We don't do that anymore, but we used to. So ⁓ those were purpose-built applications where David (11:48) Mm-hmm. Vatsal Shah (12:11) change management was heavily enforced. You built something, you don't touch it for next years, next five years. You don't upgrade it unless you are forced to. So ⁓ with the next generation of consumers of that data, they are AI agents. They evolve every single day. They get a new brain every single day. They are improving their behavior. They are following the eval sets like a human being does. So it's a new dynamic. The consumers are evolving. The platform is one of the approach where if platform is ever evolving, the consumers of those apps will ever evolve. David (12:45) Yeah. Yeah, absolutely, absolutely. In the in the book and you are one of the early reviews. Thank you very much for doing so, Vatsal. But we we we described it as a move from ⁓ let's say yeah classical yeah classical reporting, classical BI to AI infused not even just only using AI models, but really this AI infused type of ⁓ yeah, reporting, dashboarding, ⁓ calculations, ⁓ you name it. Vatsal Shah (13:17) Yeah, exactly. And once again, ⁓ there is a difference between ⁓ generic AI or ⁓ classical AI, as you describe it, all the way to enterprise AI. And then industrial AI is further more difficult because there is so much context which is lost in systems, in silos of those systems itself. Unless you figure out how you're going to harmonize all of it. ⁓ David (13:35) Mm-hmm. Mm-hmm. Vatsal Shah (13:47) Pretty much everything is garbage in garbage out problem. And I'm actually very proud to say this, prior to Litmus data catalog, the one shorting the analytics, which is you ask AI, and with the data itself, would have been what, like 85 % accurate? With catalog itself, we are reaching like 97 % accuracy. That's just mind blowing. David (13:50) Yeah. What walk me through that. Like what what A was what is data catalog? And and B like why does that change the accuracy? Vatsal Shah (14:18) Yeah. Yeah. So data catalogs, they are not a new concept. It has been there in the enterprise world for a while. So IT data catalogs like databases and Kafka and everything else, they catalog it so they can govern it. But operational data catalog did not exist prior to Litmus. We are the first company to ever create this. What does this mean is if I ask our ⁓ customer CIO, if I ask the VP of IT, if I ask the compliance system, David (14:38) Mm-hmm. Vatsal Shah (14:53) How many SCADA do you have? I don't know. How many PLCs do you have? I have no idea. ⁓ And that question gets answered much quickly if they have more than 50 sites, more than 100 sites. I really don't know what's going on on the plant flow. I don't know. PLCs, DCS system, robotics system, CNC machines that I have, what versions they are running, asset performance management, and EAM system, they solve a chunk of problem. But they were not solving the problem of data asset. What Litmus created was, can we just create the inventory of every data asset on the plant floor? Every data logger, every historian, every database, every CNC machine, everything, just listed on a single user interface? So today, humans are going to look at it for compliance purposes, governance purposes. At the same time, AI agents are using that as a reference for all the data that they are ingesting. David (15:49) Yeah. Vatsal Shah (15:50) So the catalog is a concept to harmonize all the metadata or context on the plant flow. Catalog is one of the way. Other companies, might create the ontologies. Other companies might create the graph engine and everything else. There are multiple different ways cataloging our mind was the most efficient way to get the context out. David (16:12) Mm-hmm. However, like the the fastest way, if I if I enter a plant and I've a if I talk to the plant manager and I and I and I go like, so I have this idea around data governance, and I think I wanna, and then he interrupts me and shows me the door, because they would just go like David, I have I have other problems. Vatsal Shah (16:31) Yeah. Yeah, yeah, yeah. And that's just a reality. So the first job that I had, ⁓ there was a guy on his desktop, ⁓ the Windows XP, and there was a folder on the desktop called Passwords. And there was inside that there was a Notepad file for username and password of SQL Server. That was the governance. It's like, I own this data. David (16:56) Uh-huh. Vatsal Shah (17:00) This data belongs to me on the plant floor and whatever you want to do. I'm gonna give you username and password Those days are gone. Like right now every single data point generated belongs to enterprise because that's that's a value creating asset. It's an asset so ⁓ corporate teams enterprise teams central governance team, they really want visibility of what exists if you don't have visibility you can't improve so Catalog was one of the most efficient way to surface all the metadata from the plant floor to AI and more. That's how we built it. David (17:42) But that's interesting. And it's it's indeed it's it's another category. ⁓ and it's definitely it's definitely or it was definitely avoiding the market. ⁓ let's let's go back to that multi-trillion dollar labor. actually this ⁓ just this week, it's it's funny, just this week I was so the Vatsal Shah (17:50) Yeah. Yeah. Yeah. David (18:09) The the the book is highly ⁓ like there are a lot of concepts which are influenced by the ⁓ DevOps movements in IT. We we do that because we believe that there is a lot of that there are a lot of things to learn from what IT went through more than a decade ago. We also say that it's not just a one-on-one match, yeah. So obviously it it it it's not that easy. ⁓ Vatsal Shah (18:16) Got it. Yeah. Yes. David (18:32) But I saw I was talking to somebody who's living in the pure IT world, so finance, HR, you know, that type of ⁓ things, no manufacturing. ⁓ and he he asked me the question, like, David, like in the sectors I'm working in, ⁓ we see a a huge impact of AI, not just in productivity, but in in people getting laid laid off. we hear that all over ⁓ the place. ⁓ people's jobs get replaced by Rather, that's for a good, like there are good and bad examples, as always. But like he asked me the question, and I think it was a very good one. He said, like, What what is the impact you're seeing in manufacturing today? And I said, like, the thing or the difference, and I I wanted I wanna get your opinion on that as well, Watsall. The difference I'm seeing is that today we typically don't even find enough skilled operators and we don't even find enough skilled technicians. So at this point in time, it's not about replacing highly educated operators or highly, highly skilled technicians. It's about with those who we have, like and especially with everybody who's retiring, just being able to keep on running the plant and b, hopefully run the plant better. Vatsal Shah (19:51) Yeah, no, you're spot on. in no shape or form, AI can replace human workers right now. It is not possible. Like with the current technology advances, it's not possible right now. So let's think one step further. 20 % of white collar workers, are doing the job that they were not hired to do. ⁓ So plant operator, somebody's full-time job is to maintain the operations of plant. But now they are creating reports for compliances, report for ESG, and they were not hired to do any of that. They are learning on the job and they are evolving it. Let me take a real example just from earlier this week's discussion. Litmus is getting quite aggressively getting deployed in pharma industry right now. And that was a huge uptick for us this year. I did a quarterly business review with one of the C-level executives. And we were talking about advances and where Litmus is going over the next few years. And I proposed them this idea that we are creating this framework where things, they will get augmented. Like he started, I asked the question, where do your people get burned out on? What would you like to eliminate? on the plant floor that everybody hates doing, but they have to do it anyways. So they told in regulated industries, documentation is very critical. Documentation is ⁓ like the changes that you did, recipes that you altered or recipes just you augmented or a batch that you changed, quality that you fixed. There's a lot of documentation that gets generated. David (21:36) Mm-hmm. Vatsal Shah (21:48) Can you create agent which is consuming data from Litmassage that is continuously doing these things on the pharma industry? I would give you as much money as you want. And every single person on the plant folder will give you as much money as you want just to get that problem solved. then we opened up a ⁓ like our agent marketplace, and we opened up the documentation agent. And they're like, shut up and take my money. This is exactly what we are looking for. David (22:05) Yeah. Who Vatsal Shah (22:17) I was like, it doesn't work yet. It doesn't work to the 30, 40 % that we want to, because you don't have data, you don't have context, you don't have everything integrated yet. So let's solve that problem. So when we turn this one on, it is auditor ready documentation and instantly the project got accelerated. So it's not about replacing humans. David (22:24) Mm. It the this yeah, I I I love the example. with the risk of being a bit black and white here. ⁓ that that's ⁓ that reminds me of the current let's say ⁓ problem many people are facing is you're getting an email which is clearly AI generated and it's it's like that long. ⁓ and and instead of a short list of Vatsal Shah (22:55) Yeah. Yeah. David (23:00) do this, do this, do this, or I saw this or this or this. You get dear David, thank you so blah blah blah blah blah blah blah blah blah blah blah. And then I'm using my my AI to summarize that back to the essence. ⁓ I I I I do believe like there has to be also for ⁓ validation in related industries. ⁓ you're talking about pharma, let's talk about the car manufacturing industry. Vatsal Shah (23:05) Yeah. Yep. Yeah. Yep. Sure. Yeah. David (23:30) well manufacturing, the let's say car developments, ⁓ RD, where ⁓ you know we're we're trying to ⁓ prove to like every single governmental body that's the the car we created is actually you it can safely drive on the road. So that also comes with a ton of documentation, which I'm I I hope these companies are using more and more AI to do so. Vatsal Shah (23:46) Yeah, yeah. Correct. Yeah, or at least augment the human being so they don't have to go and learn every single one of that element. But you're going to learn over 10 years of the job. But at day zero, we don't want to scare the young graduate that decided to join aerospace that hit this 300 page of documentation before you can do your job. So there has to be a better way. We want to attract young talents to join manufacturing. David (24:12) Mm-hmm. Vatsal Shah (24:24) What can you do that doesn't really scare them to join manufacturing? So you have to augment them. You have to give them whatever they are looking for in a way that is Instagram ready. Like, I'm going to swipe and I'm going to learn a new code for FDA. Can you do that? Can you make it so simple that people can get used to it? If answer is yes, you are winning. David (24:45) No, absolutely. Like ⁓ Can I ask you a crystal ball statement? Or so so let's let's let's assume ⁓ that we're continuing our the the the the path we're currently on for the last two or so years. ⁓ like what will happen in Vatsal Shah (24:50) Shut up, sir. David (25:11) One, two, three years. What are the things where manufacturing is I'm I'm sure it's not robots walking around on the shop floor. Vatsal Shah (25:24) So ⁓ there are customers they are trying with robots, like once again, for highly repetitive things, which is pick and place or ⁓ like ⁓ unloading and offloading from the conveyor belts and all of those things, has been there for a while. ⁓ Like pick and place robots, are there like since 2005. David (25:44) Yeah, yeah. Yes, and they are s and we're not even talking about the humanoids here because people now go like ⁓ y no no no no like just super efficient pick and place robots vacuum Vatsal Shah (25:59) Yeah. And they're highly efficient, highly repetitive. The mean time to failure is significantly lower than anything else that you can find in the market. this ⁓ from three axis robot to six axis robot and everything else, like it has been solved. ⁓ And we have been using robots even before humanoid scheme. So robots are not replacing the jobs. ⁓ But at the same time, ⁓ like I was in a battery manufacturing customer. David (26:08) Hmm. Yeah. Vatsal Shah (26:29) ⁓ They can't, ⁓ there are hazardous environments, it requires intelligence to operate. It's not just pick and plays. They are experimenting with spot robot at the time and then they were experimenting with other Boston Dynamics robots, which is for hazardous environment, you don't want to send human there. Human life is more valuable than anything else in this world. So you don't want to send that person there. Might as well send robot and I would gladly pass that job full time to a robot. David (26:47) Yeah, makes sense. Vatsal Shah (27:00) if the humanoid gets capable doing this. I think humanoids, versatility and intelligence, they are going to matter more than just being a robot itself. Their way to think, their visual intelligence, ⁓ their capabilities of awareness around everything is going to matter more in my mind. David (27:20) Yeah. And if we if we ⁓ forget about the robots, like ⁓ if we just go back to data where do you wanna be in ⁓ or where do you wanna the industry to be in a year, in two years? What are the missing links? Is it is it is there is is there still a technology gap? Is it s simply an an adaptation problem? ⁓ is it like what what do you think we should do? Vatsal Shah (27:26) Yeah. I think like we have to, like I'm CEO of one of the leading data technology company out there, but even I'm thinking we need to get out of platform mindset and start thinking outcomes mindset. This is where everybody needs to go in my mind. Like you get in front of manufacturer, all of these vendors, they were promising digital transformation. And you kept on delivering tools after tools after tools after tools to manufacturer and eventually get to a point where you might have a digital transformation. The idea would be there are some fundamental building blocks. One of the first one is you need to own your data. So you own that data at the edge or at the cloud, by all means own it. You have to document your processes in a way that next generation of human engineers or AI engineers, are going to consume that. David (28:27) Yeah. Vatsal Shah (28:46) context of it. These are non-negotiable building blocks. And the third one is cybersecurity. In the layout that we are in right now, securing your parameters, your infrastructure, securing your creating resilient infrastructure is critical. So you have this fundamental building blocks, which are human-crafted technology-enabled building blocks. are mandatory. Then you go, ⁓ OK, these technology pieces, exist. David (28:57) Yeah. Vatsal Shah (29:14) Stop fighting about that. Stop resisting it. ⁓ Stop denying the fact that you can't manage like 100 tools. Just have a solid foundational block. And then think about outcome that you want to drive. Outcome might be, want to reduce my downtime. I have a consistent quality problems from my CRM system. I have this failover. I have this callback from rollback, callback from my automotive system. So the idea would be understand the problem. and figure out if there is an agentic system, if it can help like 10%, 20%, take that and utilize one of the tool or toolbox or infrastructure to really empower that and solve the problem and take your company further. This is in my mind, it's so simple. And you know what we get instead of this one? Hey, I'm going to purchase MQTT broker and I'm going to purchase data op system. I'm going to purchase this database. And I was like, that's just like. David (30:00) Yeah. Vatsal Shah (30:14) that's really, really inefficient approach that you are purchasing tools rather than outcomes. And this is why Likmas is winning a hundred out of a hundred times, which is we get in front of this five-ender mess, we're going to win a hundred out of a hundred times because we're not talking about infrastructure, like openness and this and that. We have to talk about like for next four years, we want to drive this outcome for you. And that's David (30:21) Yeah. Yeah. I I I I've I fully I fully agree. ⁓ it's the yeah so this is a podcast, so I can't I can't I can't share anything. ⁓ but I'll I'll there there is this one graph ⁓ which I'm using ⁓ yeah quite a lot. And so on the horizontal axis there is ⁓ I always say time or money. It doesn't really matter because time and money are like it's basically the same. Vatsal Shah (30:43) in mind that's what we Yes. Yeah, time is more expensive than money by the way, but yeah, it's the same thing. ⁓ David (31:17) Actu absolutely. ⁓ and then on the vertical axis there is ⁓ what I always say the number of successful data use cases. ⁓ so in the let's say ⁓ mm I wouldn't say perfect world, but in in a in a good situation, there would be at least a linear Vatsal Shah (31:22) You're skittin'. Yeah. David (31:39) Different a linear relationship between I'm spending more time and I'm getting more outcomes. That would be the the th people often go like, so we have two engineers who are able to solve 10 problems a year. So if we hire two more, surely, surely, they are now able to solve 20 problems. But the reality is not. The reality is that, and this is when we talk about scaling, and that's why I think the concept of scaling is so important to understand, is that Typically, what we see is by connecting more systems, by having more people, you're actually slowing down because you're you're building technical depth. I'm gonna use that tool, I'm gonna use that interface, I'm gonna do that power scale power power shell script, and gonna use that FTP dump, etcetera, et cetera, et cetera. And we are basically slowing down. Vatsal Shah (32:27) ⁓ You absolutely are. I think this this is the in my mind. ⁓ It's a manufacturing approach problem. ⁓ I talked to so many startup CEOs and being one of the vendor where you don't control your destiny doesn't really help. The idea is that you solve a small piece of the puzzle. Like you can never make it impactful enough. that you are driving the digital transformation. You are at the mercy of something else that is driving value. that is, once again, purchasing point solution is manufacturing problem. It's like evergreen manufacturing problem. have like, ⁓ 1990s, even before we were born, like 1980s, 1970s, it just, I have this problem, I'm going to purchase this vendor. I have this problem, I'm going to purchase this vendor. They have been doing it over the last 50 years. David (32:58) Yeah. Yeah, ⁓ like physically going through the catalogue and going like yeah I need I need that pump Vatsal Shah (33:22) Yeah, that's a reality. But yeah, I agree with the, I actually use that diagram, not the one in your book, but the one in your blog, I use that all the time. The amount of use cases you enable, you call it productivity plateau in your book. ⁓ Like I use it as an asymmetrical return. If you don't have a platformized approach, you do not get a repeatable value because Once again, going back to fundamentals, you need to own the data that you got at the edge, at the local level, on-prem level, or at the cloud, but you need to own it. You can't give it to OEMs. And ⁓ you create this open platform where you keep on building the use cases and use cases are feeding the know-how back to the platform. When you do that, that one person is doing three, two people should be doing 10. Three people should be doing 50 use cases. It's not, it's not, it's not, you're not building the foundation again and again and again. And it's just so rudimentary to understand, but we go back to buying data ops and brokers and databases in the market. David (34:39) I think I think I'm I'm using that graph in almost every meeting ⁓ I I'm in. Yeah, it's it's it's it's it's so it's A so easy, but B as you say, so many people don't understand those those dynamics. But if you do, then you can start planning forward. You can start like, okay, but yeah, so that that that's like that's a good point. Vatsal Shah (34:45) Who am I? your name on it. ⁓ David (35:08) How do we how do we overcome this? How do we overcome ⁓ the plot the the productivity plateau and how do we step over that one? It's it's perfectly possible, but it it needs mindset, yeah. So ⁓ I w I I've never done this before, but I was thinking like I'm I'm the one who who's asking all the all the questions to you. This is also not prepared. So maybe, maybe you also have a question for me. Vatsal Shah (35:35) You Yeah. ⁓ I have two questions for you. One of the first one, first of all, congratulations on launching the book. That's not something that a lot of people can do. And it takes a of work, a lot of understanding, lot of evaluation before you can put that on the paper. yeah, three years, exactly. You talked about the IT OT convergence. David (35:57) Three yes. Three yes. Vatsal Shah (36:06) like in a couple of chapters quite extensively, what is your definition? Why, like how do you think it will happen? And is it already happening? Is it fused? Like the idea really is this operational technology and IT technology needs to be one technology and people need to be one people. Like processes has to be very similar. So in your mind, how far are we from two convergence? What are you saying? David (36:25) Mm-hmm. It's it's it's it's a good question. ⁓ and ⁓ so first of all, what I describe is there is a difference between ⁓ technical convergence and organizational convergence. And in the ideal world, obviously, while we are technically converging, we are also organizationally converging. But in the in I would say in reality, what you see today in most manufacturing plans, especially the larger ones, I won't say the very small companies, but especially the larger ones. ⁓ what you do see is there is a tendency to do to go to technical conversion. So we are adopting cloud technologies, we're adopting new data ops technologies, we're, you know, ⁓ We we don't want to use that Windows ⁓ machine anymore, but we maybe wanna use a Linux machine, run Kubernetes, all the fun stuff. Yeah, especially now with all the AI stuff where it's almost a given that that that you need to technically converge. Where do you technically converge? Well, at the level of the data ⁓ level of the industrial data platform, typically. Yeah, you're not really replacing. Yeah, of course, also your MES systems and your SCADA systems are a little bit changing, but but Vatsal Shah (37:27) Yes, true. New infrastructure stuff. David (37:49) The convergence really takes place on the level of the data platform. Having said that, if we are not able to solve the organizational side as well, we'll never be able to make the speeds in those changes the way we want it to have, or the way we want it to happen. And I think this is really, really important. And this is for me a very important learning from ⁓ the the the DevOps movement in IT, which happened more than a decade ago. ⁓ and for for so for those who don't know it, I always suggest to read the Phoenix project written by Gene Kim et al. And I know I know you also know that book, Vatsal. ⁓ it's a brilliant, brilliant. And you know why I wanna stress that again? Is because it's a it's a story of Vatsal Shah (38:37) Yep. David (38:44) On on on ⁓ on DevOps, but the background is it's it's it's happening in a manufacturing plant. ⁓ so that's what makes it very, very interesting. But what what happened there is IT wasn't able to evolve for the very simple reason that you had people who were developing something and then throwing it over the wall ⁓ to people who should take that into production. Now Vatsal Shah (38:51) Yeah David (39:12) The problem we are facing in IT versus OT is not entirely the same, but the fundaments are ⁓ I would say there are a lot there are more similarities ⁓ than differences to be found. If we are unable to find collaborative ways to really work together as one team between IT and OT, we'll never ever ever be able to find the speeds we wanna find because we will keep on throwing a problem over the wall into in the two directions. Vatsal Shah (39:41) Correct. Yeah. David (39:42) That's it. ⁓ so in the book we give a lot of some some recipes, some things to start. We're not saying become one team right now because it's too soon. But we end the book with an outlook where we go like, There might be something in the future, some time, where boundaries might disappear. Question mark. Maybe that's something for book two. Vatsal Shah (40:05) Yeah, sure. And I think like a few of our, ⁓ what's it, tens of our, the most successful customer, they took the talent from IT, they took the talent from OT, highly motivated talent, and they put them under the same team. They called themselves Industry 4.0 team, Data Transformation team, sometimes Data Analytics team, but they created this layout where somebody knows how operational technology evolves and somebody knows how IT technologies are evolving at the same time. Putting them together, it accelerates pretty much everything that you are doing. And that's organizational approach, one of the organizational approach that is going on. David (40:44) Yeah. No, absolutely. And it's it's one of the patterns we describe. There are also some some pitfalls with that approach. And the the most important pitfall of that approach, Vatsal, is that if you're not careful, that team can become a third silo. But there are ways around it. There are ways around it. We just you you just need to make sure that it's not it's an OT, an IT OT, or an IT problem. Vatsal Shah (41:11) Yeah, exactly. No, no, that's a spot on ⁓ approach. And I think just ⁓ in the day and age where ⁓ every single technology is evolving rapidly, ⁓ your adaptability matters more than anything else. So today you purchase something and tomorrow you might actually have to dump it because ⁓ that is not viable for a quarter later or a year later. ⁓ And how? adaptable are those teams to the new change and are you penalize them for the mistakes that they're making or are you taking that as an organizational learning and keep moving further? All of those things, they matter more than anything else. In IT or at least in Silicon Valley that I see or anywhere else, mistakes are forgiven. And in the operational manufacturing, mistakes are... ⁓ Huge career pitfall. an organizational change is critical that you allow people to experiment. You allow people to fail. You build out that organizational know-how on how not to fail again. that's just, it's beautiful if it happens. But yeah, convergence is something that we are all hoping for more and more. David (42:11) Mm. I think that's a perfect way to ⁓ or a perfect statement to to end this episode. we all we can as always, you know, we the we we could continue for this is actually this now starts to become a ⁓ a bar type of question, you know, where somewhere in a hotel lobby at the evening. Vatsal Shah (42:40) Yep. Yeah. And those are the best podcasts because nothing is scripted. I didn't even know the questions from you. And it's just our understanding of the market and bringing it to everyone. David (43:03) Yeah. No. Vatsal, thank you so much for joining. ⁓ also, of course, to Litmus ⁓ well the company for for sponsoring the episode. we talked about a lot of things. ⁓ in the article which comes with this podcast, you'll ⁓ the the listener will be able to find ⁓ more information ⁓ and also direct links to some more content ⁓ you already created. Vatsal Shah (43:13) Yeah. ⁓ David (43:38) especially or for example, around the data catalog, which is I I I encourage people to just check it out. It's interesting to know more about it. ⁓ we talked about our book, the ITOT handbook, ⁓ which you can find, ⁓ the reader can find or the listener can find that at ⁓ itotbook.com. ⁓ we're almost at the publication date, it's getting so closer and closer. So super interesting. ⁓ and ⁓ yeah, with that, thank you so much again. ⁓ and until we meet again, take care. Bye bye. Vatsal Shah (44:12) Thanks a lot, David.