From Tribal Knowledge to Industrial Intelligence: Using AI to Scope, Sell & Support Complex Solution

Welcome to the Industry 4.0 Club's fireside chat. The club's purpose is to engage the power of diverse worldwide talent to accelerate the global evolution to Industry 4.0. We accomplish this by hosting and supporting thoughtful, respectful, and interdisciplinary open conversations and creating quality content, wide-ranging Industry 4.0 topics to expand the reach of our community. My name is Mike Ungar, and I'm the CEO of the Industry 4.0 Club. I have over 40 years of manufacturing and leadership experience, most of that with Michelin. I'm joined today by my co-host, Mike Yost. Mike is chairman of the club. Mike, would you mind introducing yourself further?

Sure, sounds great. Thank you, Mike. Welcome everybody. As Mike said, my name is Mike Yost. I'm his co-host here, one of the co-founders of the club. I've been in industry almost 40 years myself, a good bit of time on the industrial solution provider side, and the past 16 years or so working with nonprofits. And so really excited for our conversations today. We are going to be talking about Going from tribal knowledge to industrial intelligence, using AI to scope, sell, and support complex industrial solutions. So we have the pleasure today to be talking with Faye Goldstein, who's co-founder and CEO of Barden. Barden builds AI decisions infrastructure for industrial automation, giving OEMs, distributors, and systems integrators an always-on application engineer to help scope, sell, and support complex industrial solutions. Faye has worked in and around tech startups for over 6 years, and we're excited to welcome her here and learn from her and have a good exchange. So welcome, Faye.

Thank you. Thank you. Yeah, my career in industrial started a little later than you folks, but it started when I was in college working in an automotive condenser and compressor shop. So I actually, you know, had a little— got my hands a little bit dirty back in the day. And then later on in technology for the automotive space, actually, for with telematics data. So a little behind you folks, but get in there.

Oh no, no, no, not behind.

Yeah, we all bring something different to the table. That's what's great about these conversations. So we're glad you could join us. So just let's jump right in. You know, a lot of folks we talked about, or you talked about, as far as this conversation goes, a lot of manufacturers and solution providers, they rely heavily on the experience of a few key experts. And I think this is something that, that can be pretty problematic for them if they're not careful. So talk for a second about the risks that organizations face when they happen to, to deal with critical knowledge that lives in just a few people's heads instead of being more captured and available for others and shared systematically.

Yes, I think even looking at it from the industry perspective, we have right now a Massive reliance on very intelligent people in this space, right? When you're selling an industrial system or solution, you can't just rely on the, you know, the guy or the gal that worked at the grocery store to go ahead and do that. You need folks that are actually engineers that know what they're doing, that know what they're talking about, that have been involved in that space. And we're speaking now in the industrial world about an 85% deficit of these engineers, yet almost 70% of the industrial systems that are built, you actually need folks like this to actually go out there and scope and sell them. So we're talking about, in general, not only are you relying on a few key experts, you're relying on a few key experts that soon won't really exist, that they won't be there to engage with you and to do the systems that they— and build out these systems that they actually need to be building. Relying on a few key experts that actually do exist, the ones that are here, we're seeing roughly 50% of folks in the space retiring, right? So we're also talking about massive amounts of folks that are leaving that have the decades of experience that are leaving the industry or moving towards nonprofits or building out the ecosystem in other ways. And then you don't even have enough folks coming in. So where, where, who, where's the information being left? It's being left in these heads. And it's a big issue in order to expand the industry.

So are you seeing that the, I guess, the deficit, is it typically on like the end user side of these types of complex systems, or is it more in the supporting organizations, the systems integrators and the consulting firms? And, you know, where do you see, or is it all the way across, I guess?

So the 85, yeah. I mean, the 85% number is broad in the manufacturing space, right? So it's kind of a broad number. We're talking about engineers that are going into the actual, on the manufacturing floor, those are the engineers that are going out and scoping the projects. Those are also the ones that are actually building the manufacturers themselves. So it's on the end user side, the supplier side, the distributor side, the integrator side. We're talking about a broad global, I would say, number when we're looking at things. And I think that we kind of forget how often and how many times within the value chain, you have these folks involved in the process, right? You have engineers involved from the pre-scoping stage all the way to the end when you're trying to support or troubleshoot. And so that deficit is massive. And precisely where it's the heaviest, I think the problem is that it's burdening the customers, and that's where the impact is felt the widest, right? It's not even about how many there are, how many there aren't, is the fact that customers wait longer to get their projects scoped. They're waiting longer to get their systems supported. The folks that need robots on their manufacturing floor can't get those deployed to them and working fast enough. So the big impact is the impact on the customer side.

Right, for sure. Well, so one of the challenges obviously is how do you scope, you know, the complex customer requirements, right, in the first place? So how, are you seeing? And I guess, are you seeing it actually happen or are you seeing the opportunity for AI to really come alongside and help these organizations translate their needs into, you know, the technically sound solutions while you're actually less reliant upon these resources that you're talking about?

So I would say kind of a mix. First, I think we should look a little bit about what does scoping an industrial project mean? Oftentimes we're talking about pre-PO phase, right? They're not getting a paid order.

Mm-hmm.

You have an opportunity, an RFQ comes in, and you have massive amounts of requirements that need to go, again, up to check of electrical, mechanical engineering, systems engineering, chemical engineering sometimes. And the big question is, can we even build a system that solves these customer requirements? And so essentially what you end up needing to have is you need someone, somebody, some knowledge that sits in between everything the customer needs to to be able to analyze those requests, break that RFQ down, understand what they need, and then everything that sits also facing the product and the solution builders in the actual supplier side, the people that are deploying these automation systems, to actually say, where can we meet in the middle? How can we engage with this? And so when we're looking at the scoping, there's many different parts that go into a scope. From the beginning part, it's even breaking down and understanding what is the customer even asking me. To understanding what are the requirements, to understanding what are the standards I need to comply to. And essentially, one of the things that we've spent the most time doing in, in Bardon is breaking down the decision-making pathway that an industrial application engineer takes in order to both scope projects from A to Z and to then support those deployments afterwards. Going down into the nitty-gritty detail, I don't know if we have time for that. But there's so much that goes into it.

Sure.

Well, and it makes me— when you describe that, I'm thinking about the technical solutions I was involved in in my day. And the issue wasn't getting a piece of equipment onto the floor that worked. It was actually having the— having the people that provide that equipment have the knowledge to create and implement a solution that had good Maintainability. In other words, they, they, they had— they didn't have any idea about how to make it cost-effective afterwards installed. And I can imagine capturing all that information and having it really be built in so we're not worrying about, did we tell them that, you know, we got to be able to reach that, that, that device without having to take the machine apart, you know?

So, so one of the things that's incredible with AI, I would say, is that AI, obviously, people a lot of times think about it as an LLM, right? Large language model that can do predictive thoughts about what should come next. If you're building it right, however, you can actually fine-tune or train AI to then give deterministic answers. What does deterministic answer mean? Means that the AI is not just looking at its entire mass of knowledge that it has from the entire world, but if you're structuring a knowledge base for it and saying, here's how I need you to look at this question. Question or look at these requirements, look back at my historical ways of doing it. Now, if you're thinking about industrial, these system integrators, these suppliers, these many OEMs that are selling these systems have been doing it for 10 years. They have years of history of successful implementation that humans have led. So now there's a massive opportunity to say, how can we learn from— and if you're looking the way we're thinking about it, the delta between these initial RFQs that have come in maybe even 5 years ago The end scope and BOM that you built, let's say, or the bill of materials that you built, or the end proposal that you submitted them, and then the end quote that you gave. If you're looking at kind of the middle ground between these things, that's where a lot of the knowledge and the decision-making sits. What did they need? What did we end up giving them? How can we— and this is what we're spending a lot of time doing— how can we map that out? How can we map that historical efforts out in order to then help the AI make a future decision based on past historical successes or failures so that it goes ahead and makes the right ones in the future.

So I'm wondering if I— so I grew up in the automotive tier 1 space and on the solution provider side. And one of the, I don't know, one of the things that we realized that we said all the time is that everything changes but the launch date.

Right.

So as we're trying to scope things, and really that was a reflection from where we sat of trying to build something and then it moving and moving based off things we learned or things that customers learned. So I'm wondering that knowledge that you're talking about there and capturing that, what tools that you see or how much you think that that can help. I guess it's really a problem on both sides, right, of of understanding what it is that we're actually trying to build and then, oh, that wasn't quite right. And, you know, how do we improve that process overall?

Yeah, I mean, I would say sometimes what we're hearing from some of our customers is that even knowing how to ask the right questions in a sequence of asking the right questions to the customers is helpful to understand how to not make initial mistakes in the scope. So if we're looking at this, let's say, initial need of a customer, let's say we need a new— you've got your Pfizer and you need a new conveyor system that can actually go and drop the new medicine into glass bottles moving at a conveyor at this and this and this speed. You need a specific sensor that can have a very specific washdown that could be FDA compliant, right? So that's a very, very broad— that's requirements. And of course, yes, it's going to change along the way, which is actually in fact why the human labor involved in this is so expensive, because you constantly need them to come back in and rethink and rethink and rethink. But one of the things that you can think about is if you're breaking down this customer requirement, Pfizer. Pfizer. You can break it down. Okay, we know the industry. We can have one element that we're understanding mapping of the industry. We're talking about a conveyor system at this specific speed. What conveyor system? What speed? We're talking about a sensor, flow sensor that needs to measure a certain type of viscosity of liquid. What is that sensor? What is the viscosity of that liquid? All of these different elements that us as a human intuitively try to put piece together essentially can be broken down into individual data points that can then be mapped. And what brilliant with AI is that you can have hundreds of millions of trillions of data points that can be mapped, that you can— the AI can find these relationships with and make an understanding of saying, well, if we're taking similar parameters for— even though it's not Pfizer, let's say we now have an automotive or a tier 1 that's going on a conveyor system at a similar speed. And even though they don't need a washdown at that exact same degree, we can still understand the lessons taken. I spoke with a company, an engineering firm the other day, And she said, you don't understand intuitively. I can understand how if I was scoping out a project for a pet food company and they were dividing the pet foods according to the colors, and I needed a system to make sure that the colors was going into the right bags, I could apply that knowledge to a system that I was doing for a pharma company that needed to divide the pills into different coloring boxes. Humans know how to do that, but to train AI to do that, that's where you need players in the game like us.

That's cool. Wow.

Did that answer your question, Mike? I don't know if I went on a tangent or if that touched really what you were asking.

Yeah, I mean, it's just, it's a dynamic space that we're in, right? And anybody who's ever lived through a go-live project, especially automotives and pharma and probably any industry really, but I mean, there's a lot of pressure there. There's a lot of just dynamic things that happen. Constantly you're trying to respond to. So the idea of having, you know, a way to tap into the experiences of others that have gone through such things is really a pretty compelling argument to say, yeah, we can avoid some of the skinned knees along the way.

Yeah, I mean, I would even go farther to say, if we're looking at the US market and kind of the way for the US to win, it's really this high-mix, low-volume opportunity that we have, where you have a constant change in how are we building things, how are we going to make this change for this specific very niche application or this specific defense element or this specific aerospace need or this specific thing. We're talking about very, very complex builds, not a lot of them. So that repetitive nature that typically folks see in global manufacturing, I don't believe is the way for industrialists in the US to necessarily win. We need a way for us to be able to say, how can we deal with this very high mix, low volume complexity? Complexity that is constantly changing over time. Every new input could completely change the way the output should look. As humans, if you're really, really, really trained, yes, you could understand and you'd be able to say, oh, wow, they actually added in a different robotic arm that actually speaks to a different type of PLC. I now need to completely change the system I built because it no longer speaks to the same data. And has required humans to be very intuitive about those things. But if you're building AI to do it, you'd be able to help that process a lot.

Yeah, but just one quick aside, Mike, if you don't mind. An analogy I've used, or a parallel that I've used when I've talked about the kind of reality that you're talking about, Fei, is that I know growing up as an industrial automation guy that I could have built a cashierless checkout system at a grocery store. Right? But I would have had to hardwire everything from what product is in what bin and how do you load things in, how do you take things out, how do you handle error-proofing and things like that. And now in the age of AI, we have cameras that learn what people are taking and you don't have to reprogram everything when you change the products that are on the end cap. So I think that the power of AI today versus, you know, the stuff I used to sell years ago certainly lends itself to This kind of thinking. And I guess that's what probably excites me the most, is that we can really start to challenge this thinking because we can do things today that we could have done before, but not as quickly, not as reliably, not as cost-effectively.

So yeah, I would like to say that if you look at AI as building it in a way that could take kind of the neural network of a way a human thinks, and at least the way I think, it's like I go to A and then I realize B and then I realize C, and you can even hear probably in the way I speak My brain is running at a million miles an hour, and there's so many different connections that we almost intuitively make when we're making a daily decision. We think about 5 steps ahead and then go backwards 1 step and change direction. Now, it's complex to assume that we can build a system that can almost replicate that, but we're getting closer. How do you—

following that question, though, how do you make sure that you have a deterministic model? Versus the probabilistic model. In other words, you're going to, you know, we talk about all these different variations you could possibly build and all this knowledge we're going to have. And you just hit it on the head as a human. I can— a lot of that's intuitive to me to sort through the little nuances that are unique to this situation. How do you handle that in AI? Because that's not so easy, I think.

Yeah, I think the trick is for me is I'm— we're not saying we're making a deterministic model or a, you know, probabilistic model itself. It's what is the data that it has access to and what is the orchestration we're doing behind the scenes that before the model gives the answer, before there is an answer from the LLM, it runs through an orchestration that is deterministic. For example, you have a choice of how to build the database. Are you building a classic semantic search database? Are you building out a database that you just are looking at a whole big body set of data and then choosing which ones make sense to have as they're the highest ranking probability? Or are you building in the way that we're building it? We're building a knowledge graph, meaning we're having actually very, very hard, what are called kind of nodes of information, and there's strong sticky relationships between these pieces of information. So if you are now having an AI model go and say, don't just look at all your knowledge of LLM, don't act as an LLM, but look at the database and understand and follow relationships. of where that data was in the past and follow these relationships between these data points, you then enable a more deterministic decision or a more deterministic answer. Now, to be frank, I don't know if we're going to yet in the next little while get to a fully 100% deterministic answer. I think there's almost a beauty in having certain things that are LLM-based, that are large language model, that are not just deterministic because deterministic would just be machine learning, right? That would just be, classic building a taxonomy tree and building a tree of saying, if A, then B, if B, then C, if C, then D. The beauty of AI is that you can mix that with this fluidity that typically comes with humans. So I think having a desire to be fully deterministic is not necessarily the best thing for solving these problems sometimes.

Right.

I think you need a high degree of determinism and auditability of those AI answers, and auditability is an important thing, But you also need to be able to understand if we really want this to be my human copilot or my human coworker or augment me as a human, it needs some free thinking opportunities. And so I think finding the balance is important.

Yeah, I think that's what you're talking about there. And we started— we talk about it becoming a competitive advantage, you know, and the productivity piece, you know, I'm getting— you're talking about the AI doing all the things that that are routine for us already. Is that what we're talking about here?

So I wouldn't say that it's necessarily routine because routine in our day could be tying a shoe and I don't expect an AI to come and tie a shoe. What I actually mean is the knowledge work that is taxing to a lot of folks. So I think that's really where the magic comes into play when it comes to the AI that I'm speaking about, not the routine things, but really the draining knowledge work that these folks, people need to be doing or the complex knowledge work that typically takes 5 people that can now be done with 1 person. Or in our case, right, we now can enable a younger, smaller integrator or a smaller component manufacturer or smaller, newer robots as a service company or robotics company to actually be able to scope and deploy more projects because it takes a lot of time to train these engineers in order to do all of these things. And so now you're enabling the smaller folks to compete with the bigger folks because they now have a capacity that's not completely tightened. They have bigger capacity to do more work, do more things with the same amount of people that they've got without spending the time to train a whole bunch of new folks that don't even really exist to train.

If I could just remind everybody here that we are speaking with Faye Goldstein, who is the co-founder and CEO of Barden AI. We're talking about using AI to scope, sell, and support complex industrial solutions. And so Really appreciate the conversation, the energy so far. I guess one of the things that goes through my mind, Faye, especially when we're talking about learning from people, is who are the people? And especially in a complex system where you're working across— you're working with a manufacturer, you're working with, you know, potentially a technology solution provider, be it robotics or software or whatever it is, you're working with systems integrators, et cetera, the knowledge is in a bunch of different organizations, right? So who owns the knowledge, right? Or who's contributing to the systems that are learning together? Is that the Pfizer in your earlier example, or is it the individual folks who are part of the ecosystem?

Yes, that's a great question. I think that's one of the biggest, I guess, detriments to the speed of AI implementation in the industrial space, obviously, is the fact that the data is so siloed, it's in so many different places. Sometimes it's in email, sometimes it's in SharePoint, sometimes it's in your ERP, sometimes it's your MES, sometimes it's in a PLM, PDM, wherever. We've got the alphabet soup of all these places data can sit and live. Our focus is to focus on the data that are coming from the sales and support team of the sellers. We're very focused on the sales side of things. So who is selling, scoping, and supporting the systems? We are not focused on the procurement side. Now that's a whole nother big problem to solve, Not the problem I'm looking to solve right now.

Yeah.

Our belief and the way we're thinking about it is you can have tons of procurement agents or tons of procurement AI tools trying to procure an automation system. But until all of this information from the seller side is actually ready and machine-readable for any procurement agent to actually make a proper educated decision, these procurement agents will never be able to buy a system. They might be able to buy spare parts, they might be able to buy replacement, but to buy a system, no procurement agent will be able to do that. So to start and build this world where commerce in the industrial space flows most efficiently, we first need to start and say, how can we take all of these different disparate pieces of data? And sometimes you're right, the data of whether or not this project works for this company is something that sits and we understand, oh, well, it's a pharmaceutical company, so therefore we need to look at these specific standards. That might sit somewhere else.

Mm-hmm.

And if it's, let's say, looking at the past projects you've built and you're saying, oh, well, what this is actually, we're building a replacement for an existing system, you need to know what exactly actually exists there in order to take, and that might look into, let's say, an old bill of materials that you built or the old drawings of the systems that you've built. There are a lot of different types of data, and I think that's why when people are asking about why now, in the past, yes, connecting into all those data sources would mean normalizing them into a very, very strict requirement. I think one of the brilliance, and this is an argument I have sometimes with folks about AI, is like everyone says garbage in, garbage out, but I don't believe that. I think that there's opportunities to find magic in every single element of data that you might have. Now, it may not be obvious and it may be within that delta, right? The data that you don't see, but just the difference between this point and this point, that might be where the information sits. But products and more and more, and I think, Mike, you know this most, is there's a desire for this market and for our world to start making these pieces of data accessible so that the users and the customers are actually being able to use it. So yes, it's still a little bit of a hurdle connecting into every data point. Now, us particularly, we connect into service tickets. We understand all the service tickets. We can connect into your old historical projects, and we can look at all of your SharePoints and your historical projects to understand how do you typically scope and then support these builds. That means looking at the bill of materials and the quotes and the RFQs, etc., etc. And then other folks need to connect into other data pieces, elements of data. But I think There is a, really right now is the time to do it, for folks to open up, to realize that if they give some access to teams that are working to solve some of their biggest labor problems or some of even their biggest knowledge problems, there's an opportunity to solve them. That's easier, way easier today than it was even 6 months ago, than it was even a year ago, than it definitely was 10 years ago.

For sure. That's neat. So, if you think about the future, take us into the future now. So I'm going to close my eyes and think, because you're talking— I mean, I really appreciate your enthusiasm for this, the thinking ahead about what things will look like in the future, could look like. So take me forward 5 years. What, what do you imagine the industrial sales engineering and customer support functions will look like when we really bring this thing to full fruition? What are you thinking? What do you think it looks like?

I believe that there will be a web of interconnectivity between everyone in an organization. So you have, let's say, Sam that is going out and he's going to meet a customer in Wisconsin. Sam, if he was in the company now today long enough, he'll know that maybe 2 years ago we did something similar to another similar company. In 5 years' time, Sam will intuitively know that. He will come prepared understanding everything that he might need to know, even from the time before his time. I do believe that people are still going to need to be building the personal relationships no matter what. I think that that's going to be the core. People are going to be hired based on their ability to engage with their users, to understand problems, to have intuitive conversations, and everything that is relying on connecting the dots behind the scenes, that's where AI will come into play. So in our future world, you'll have an RFQ that comes in from Pfizer that needs to build a new automation plant. They need to build a new conveyor system. It comes into Sam and his team. Sam already has half of that scoped already. He knows exactly the questions he needs to be asking to verify in order to make sure they're scoping it properly. He has all the learnings from the Merck build that he might have done X amount of years ago or that John, his colleague, did X amount of years ago.

Yeah.

He also will understand intuitively, well, if we actually scope this one, this product is actually at an end of life, we need to probably think about this. And I know that Pfizer likes using these type of robots versus these type of robots, so then therefore we're actually going to be putting it in. And it's not going to rely on him knowing it, He's going to have guidance through AI to know all those things because they've done it before. Someone's done part of this before, and now he has access to that piece of information. The way I'm seeing it in 5 years' time is that any company that wants to deploy and automate their— that deploy automation, or any of the teams that want to actually automate their factory, will have a hell of a lot easier time doing it with far less back and forth, back and forth, back and forth, back and forth until it's actually implemented, actually working, and actually running to make people more money.

Wow. Okay. So given that, let's talk for one more second about the capabilities leaders should be building in today. So how do we build this foundation so that we're able to accomplish what you just described?

I think number one, understand that— and this is again a big hurdle that we're seeing— is there's a lot of folks that tell themselves, we have an AI strategy, we're going to spend the next year building out some AI strategy, and what we want to do is we want to work with some of the biggest players to build AI that goes across everything in our organization, one company solving it all. And I think that's kind of a mistake, especially, especially as an industrialist, right? And Mike, you said like you could have technically as an engineer built the cashier system, but in theory you were doing something else, right? That in the same way in our world, you've got the folks that are building the robots, and yes, they could go ahead and build a conveyor system, and they could also build the motor and the motion control, but everyone's kind of staying in their lane. And so I think the big thing I would say to folks is You're gonna be too late of trying to find what really works for you if you spend the next year building strategy and you spend the next year assuming that there's gonna be one magical company that comes in and can solve it all for you, especially the complex things, especially the things that you need to make sure that this is an expert in the space. You do not want the company that is going and engineering the bicycle for your kid to do your Boeing landing gear, even though both of them have wheels. Yeah. You don't want that. And so I think that's the biggest thing that we need to be thinking about is how do we enable experts that are building for use cases to succeed with those use cases? That's number one. I would say number two is there's this big fear about sharing data and opening data. I think find trusted partners and be flexible and know that in order for them to build for you, for what you need and to solve your problems, give a little bit, give a little bit of an understanding of if I share even old stuff, Even old stuff that's not necessarily relevant, but open it up a little bit and recognize that in order to win, you need to actually release a little. And I think that's how I would say, don't worry about complex data, don't spend a year trying to build a strategy, and trust startups that are really working their butt off to solve these things for you and to be flexible with how you're engaging with them.

Very good. Well, it's been really great listening and learning from— with your expertise and just talking about this topic of building these complex systems and how AI can help us to do that in the future. As we get ready to wrap up, Mike, maybe a final thought for the audience, and then, Fei, you as well, and then we'll call it a session.

Yeah, for me, you're speaking my language. Not necessarily AI, I'm not smart enough for that, but certainly the complex industrial solutions and how they get scoped and how they get to delivered. And I, for a number of years, I used to talk about the systems integrator dilemma and the fact that we don't have enough systems integrators and that, you know, their role and their value is oftentimes— they are oftentimes undervalued. They're expected to keep, you know, knowledge in-house, expertise in-house, not only for new technology that's coming out, but potentially like ridiculously old technologies, right? You got to keep those, those systems that you put in place decades ago running. And, and, um, so they're expected to have that kind of knowledge, that kind of experience, that kind of expertise, um, and, and can be very niche things as well. So, you know, to me, you're really leaning into a way to address that situation, uh, at least being able to make those people into the role that they play, it seems to me that you're leaning into an area that I've believed for a long time is really valuable. And if you can make those people more effective, then that is fantastic. I hope that what comes along with that is an understanding of the value of how hard it has been for them for all these years to learn all the lessons that they've learned along the way and the true value that they bring. So I I think it's fantastic where you're leaning in and certainly your excitement and your vision and your focus. I hope folks will get on board and keep working this together and see where we can go with these things.

Yeah. Thank you.

Okay. How about you? Last thought for our audience?

No, I think we're already deployed, we're already working with some incredible customers. I'd be happy to speak with more folks and even just get learnings from you. You know, just sit and understand your history, how you did things, where are the pain points that you're dealing with. I'm really excited to speak to more folks about it and get our solutions out into the world even more. I also, I think that there, you know, Mike, touching on what you said is like there's this question about system integrators and the role of integrators. And I think that we have this when you're looking at the landscape, this this concept of integration sits on many, many sides of the value chain. It's not just system integrators. There's integrators on the side of the suppliers that are integrating those components. There's integrators on the side of the OEMs that are implementing these things. And even in the end user side, they might not be called integrators, but they're integrating the systems they're buying. And so it's not just the systems integrator, it's across this world of this engineering, application engineering mindset that I think is powerful to think about. And that's where we're focusing. I love chatting with people. So anyone that's listening to this afterwards and just wants to talk, wants to brainstorm, wants to have a fun conversation, please reach out. Please connect with me. I love posting on LinkedIn. So follow me there as well. And I'm excited for the future of this space a lot.

Wonderful. Thank you, Faye. So I want to thank you for taking your time out of your busy schedule, Mike, you as well, to help us talk about AI and how it can accelerate and simplify the sales process for engineers and system integrators. The club, the Industry 4.0 Club, would like to encourage you to take action. So that might be reading an article, attending a workshop, listening to one of our broadcasts, or even starting a smart manufacturing roadmap for your factory. We'd love to hear from you, hear what you're doing, so drop us a note at our website, www.industry4oclub.com. Or on our LinkedIn page and let us know what you're doing to advance the deployment of Industry 4.0. We also host conversations like this weekly, and so if you'd like to be a guest on one of our future broadcasts, please drop us that note as well, and we'll certainly explore that and see if we can have you on to talk about Industry 4.0 and, and how to help drive that into the future. With that, thank you, Feng. Thank you, Mike. I wish everybody a great day. And a good rest of the week.

Thanks, guys.

Take care.