Trace3’s Josh Lindstrom on Enterprise IT Challenges with Generative AI

This is Textron tv. Hey guys, thanks to the throw. We're here with Josh Lindstrom, who's senior director for Data and Analytics for Trace three.

And we're talking about all these AI tools that everybody's using without necessarily knowing the providence of them and exactly how they work, and perhaps that has a certain amount of cybersecurity risk attached. Hey, Josh, welcome to show. Hey, Mike, uh, great, uh, great to be here.

Uh, appreciate you having me and looking forward to this conversation. It's, uh, it's a top, uh, topic for, uh, a lot of, uh, our folks internally within Trace three and obviously with our customers. So looking forward to, uh, the conversation and seeing, uh, and, and kind of seeing where it goes.

I think a lot of the compliance people out there, the security people, and even the application owners are a little, shall we say, having a sense of adjective right now because the employees are running around going, look at what I found. This is a great new tool. It does X, Y, and Z, and it's all AI generated, and they're just kinda invoking it, throwing a bunch of data at it and not realizing where that data may wind up.

Well, caring for that matter. So, are we in some sort of kind of crazy, wild, wild west scenario that's gonna come back and invite us? I mean, I think it is a little bit.

I mean, you're, you're hearing kind of the comment around, uh, AI bubble and, and AI winter, uh, comment where I think a lot of the challenges are, uh, organizations aren't, aren't really understanding what the true ROI. Yep. You can go in and help productivity, but what does it mean, you know, downstream, right?

Um, hey, I know I gotta do something, and I know that there's aspects of, um, uh, of use cases that I gotta go after. Uh, but what does that mean for any sort of R-O-I-T-C-O? Um, all of that kind of coming into play, and then you add the layers of governance, privacy, and security across all of that.

It becomes even more challenged. And then probably the biggest thing that we're seeing, um, yeah, anybody can identify a use case, but the challenge is if you don't have good data quality, um, and you don't have governance around that data, um, that becomes another, uh, sticking point. So you're kind of, it's coming from all different angles.

The most organizations that I think that are really understanding, um, you know, from our perspective, um, they're, they're rolling up the sleeves. They're, they're, they're drilling down on enablement, uh, of their people. Um, they're working through those security vulnerabilities and knowing that that's an issue.

They're, they're laying, uh, some of that groundwork ahead of time. And I think that's really what we're seeing from our side, uh, holistically. Mm-Hmm.

So How do I bring some adult supervision to all of this? Because to your point, there are economic issues, there are compliance issues, there's governance issues, security issues, and there's stuff all up and down the stick. Well, there's, I mean, we're getting inundated with AI marketers and talkers, and I think that's where the key differentiator from, like a organization like mine in Trace three we're doers.

And so we saw this early on, and we're dealing with a lot of the Fortune 1000 companies that are really kind of trying to design this AI framework and, and all kinds of aspects of that really kind of, we, what we've said is like, look, have a good governance framework in place. Lean on that out of the gate. And if, and if you're not and you're kind of just building things in, let's call it shadow it, and now you've got something that actually kind of comes outta results, well, what ends up happening is all the business folks on that side start raising their hands saying, I want that.

I want to be able to lean on that, um, and bring that to fruition. Ideally, then that turns into something that now we're using into production. And there wasn't any sort of, we, we haven't identified the security vulnerabilities, like all of the things that kind of come into play here, uh, compliance, regulatory issues, all of that.

We're just using something that was built and is creating a result. Um, and so what we're saying is like, take a step back, understand that we have several organizations that are deploying copilot, for instance, what is the governance around copilot? Are, are you opening it up to everybody or are we shutting it down so that if I'm leveraging copilot, I can't go ask, uh, how much our CEO makes in their salary, right?

And be able to pull that information. So there's all kinds of aspect, and we really feel like good data engineering on the front end creates good data quality, creates good governance. Um, we're seeing a lot of organizations struggle with that piece, and we're helping them drive that.

So I would say, you know, put, put together that governance framework really adhere to it. Um, make sure you're leaning on the right folks to be able to make sure that the right data, uh, and you're getting good quality and governance with your data. And then probably the last thing is that continue to enable and train your team to be able to accelerate, uh, some of these things.

The more you're, you're opening it up, um, we're seeing a lot of, you know, fortune 1000 organizations do, do AI in silos because hey, we're not, we're not sure how much it's gonna be able to do, but the more you embrace the team, the more that you're sharing some of these wins and, and some of that type of stuff. Now you f put that framework around compliance and governance in play. Now everybody is adhering to that.

Right. Well, let me ask you this question then. Does all this AI stuff, is it just really exposing a lot of flawed data management practices that have been around for decades that we've kind been ignoring?

And I always felt that, you know, when I talk to companies, none of them would get a really passing grade for the way they manage data. And is all that coming home to roost then? Well, I think, uh, you know, it's interesting, we talked to so many organizations that are kind of doing a bunch of different things, right?

And it's like, well, if I create a data platform, then hey, I'm good, right? Um, and what we're seeing is that, you know, a lot of folks hurried up to go build out those modern data platforms, migrate off of their, their legacy type of platforms, and it brings a lot of ca capability. But to your point, Mike, you're dead on.

Like, we're seeing lack of data management on the front end because again, they're getting inundated, uh, by all these different data sources and being able to bring through it, right? Again, it goes back to, you know, are you, you know, have you prioritized those data sources? We're talking a lot of customers that are got a bunch of use cases, but are you ideating on those use cases?

Are you leverage, you know, are you, are you digging that layer deeper into those use cases and understanding what sort of data you're gonna need in order to go solve that use case? Hey, I wanna do, I wanna do anomaly detection on some devices. Um, well, do you have timestamp data?

Do you have the right data to be able to go allow you to do anomaly detection? You know, certain things like that are kind of coming to fruition and really goes back to, do we have access to the right data? Can we go work through that?

And we're seeing real true hand, um, lack of data management, lack of data governance. I think one of the other aspects of it is we saw a lot of organizations that built data platforms, but then kick the can when it came to governance and governance tools. Hey, we'll get to that in year two or year three of the platform.

And now you're starting to see organizations that have a platform and then are hurrying up to be able to go do it. Now, what's awesome is some of the data platform organizations are building data management as part of their platform. They're building, um, the ability to have, uh, catalogs and, and lineage inside of the platforms.

And you're gonna start to see that happen real time and then really making it open source so that a lot of people can, um, contribute and bring ideas to the table so that you're, you're adhering to that. But yeah, Mike, I, I would say that most of my customers and our teams that are talking to these folks, um, are struggling with that. Like, we're not, we're never walking into a, uh, a data greenfield.

It's always a data brownfield and they're doing something, but what are they doing and have they been, and do they have the right mechanisms in place around an overall governance and strategy, uh, that will adhere to something like AI that comes up and now we have to accelerate even that faster? Yeah. I feel like we're having this classic build versus buy conversation over again.

And if, if I buy, I don't really know what data was used to train the AI model, and if I build, 'cause I have more control over their access to the data. I don't have the tools to manage the data in the first place. So I'm kind of damned if I do and damned if I didn't.

For sure. Yeah. Is it, strike you as somewhat ironic that at least when I look at it, uh, for years we've been moving stuff into the cloud, but most of the data is on premise, and now that we're trying to apply AI models, we're bringing the compute in the form of the AI model back to where the data sits, which is largely on premise.

Sure. Yeah. I mean, I, I I would say that a lot of folks are bringing it back, right?

Um, there's still a lot of organizations that are leaning on the cloud, but you know, a lot of the c organizations that we're talking to is, you know, lamb the data, train it on-prem and make it burstable to the cloud. It helps overall cost, it takes care of CapEx versus OPEX options. It allows us to be able to be more granular on governance and, and we're not pushing everything, uh, to the cloud.

But a lot of our organizations that we're talking to, especially when it comes to data, have a hybrid approach, right? It's not one tool to rule them all. Um, they're, they're leaning on several tools, and again, really based on the use case, um, we kind of pride ourselves in the same mindset.

Like we're, we're agnostic. We're very opinionated on the technology based on the use case and what you're trying to go after. And, and again, the other aspect of that is we, we have a saying, leverage, configure your build, leverage your existing tech and configure it to work together to get the outcomes you need and only go build net new, uh, if you don't have, uh, the key components technology wise, uh, in your ecosystem to be able to get you the results that you need.

But yeah, Mike, I think to your point, you're seeing that, um, what is it, lack of customization from some of these third party tools. Uh, you're seeing integration challenges, uh, because some of it, and, and most of it is all brand new. Um, are they reliable?

Are they gonna be around for a while if I make a commitment to that third party technology? Um, are they gonna be here next year? Yeah, they may solve the problem today, but are they gonna be there?

And then going back to your comment, security, um, have, have some of these newer kind of, you know, AI technologies and tools that are in play. We know that the data platform technologies are, have been pretty stable over the last several years, and organizations are using them into production. But you're starting to see some of these third party technologies come into place and, and leveraging these tools.

Well, have they covered all the bases when it comes to security? Uh, are they adhering to all those security concerns? And I think all of those things are kind of key elements to, as you kind of start piecing, uh, some of these technologies to work together, because we, we see it all the time, um, you know, like a particular technology, Hey, Mr.

Customer, did you know, you could decrease cost inside your data platform by any, by leveraging the technology that you have on one end, uh, the right way? And by the way, that will, that will help with, uh, cost across the stack, right? And so a lot of organizations, as they kind of drive toward this AI initiative and get their data in order, now it becomes like, okay, you, we, we don't want to buy a third party tool and then try to force it into our ecosystem.

Let's see what we got. Can we go leverage that? And if the use case warrants it, well now we can go start looking at some of these third party technologies.

Do you think that the way it organizations are structured we'll need to change? Because, you know, I look at AI today and I go, wow, I need data scientists, data engineers, a couple of DevOps guys and development workers, throw in some security people, and you know, by the time I wrap it all together, or I need a small village of people to do anything, um, are we too behold in the specialists we need to find a way out of this? Kinda?

Uh, well, and I think you're seeing this, you know, giant tidal wave of transformation, and I think you're seeing the IT teams, uh, deal with that as well, right? Like, how do I take advantage of the, of, of some of the technologies to allow us to, to blend, um, you know, uh, skill sets, uh, from, from our folks in play, we're seeing a big trend, uh, in DevOps, cloud, data security really coming together and having to figure out how to collaborate the right way, right? And so, um, you know, and with our experience and 20 plus years of, of being in the mix, and I think we even internally within Trace three, we're figuring that out.

We've been, we were very siloed for a long time where our security team had their initiatives, their cloud team had their initiatives, the DevOps team did their things, and then the data and analytics team was there because, um, you know, we needed a, we needed to leverage analytics. Well, what we're seeing is really kind of that mer like organ, like the best organizations are bringing those all together. And yeah, they may be renaming them, but to your point, yeah, you're gonna have to look at, uh, it a little bit different.

Um, we have cybersecurity teams, um, or CISO looking at creating cybersecurity data lakes and leveraging AI to go do that because they're collecting a bunch of log files and a bunch of other things. Well, they have a bunch of tools that they leverage. And at the end of the day, how do I get the results?

How do I see, how do I stop breaches? How do I get ahead of, uh, you know, how do I get ahead of, uh, some of the, the challenges that I have and what are the things I wanna work at or look at from a CISO lens? At the end of the day, I wanna be able to leverage the data that's available to me to, so that I can go drive outcomes and see trends.

And so that was something where we talked to a lot of CISO and they're like, well, the data team owns the data platform. We don't even have access to that. They're starting to now become, uh, um, data citizens, I guess you can call that for a better term.

And they're learning to embrace the platforms that maybe started in the data, data and analytics team. And they're being able to leverage that to get the things that they need to get, you know, from their perspective. So On the plus side, everybody's starting to knock on the same data door, and at least they know there is a data door.

So that's progress. Um, as you kind of wonder about the projects and the number of them that can be done, uh, some people I talk to are kind of like, we're gonna let a thousand flowers bloom and hope something good happens, and other folks are, we're gonna narrow in on two or three use cases and see if we can drive those. So between those two extremes, where are you seeing more customers lean?

Are they trying to make, you know, just kind of like throwing all the seeds over the wall and hoping something happens? Or are they narrowing in on, you know, the, the two or three things that are gonna make a difference? Well, it's interesting, uh, Mike, 'cause uh, depending on the customer, you could have a hybrid of both.

And that's, that's an interesting aspect, you know, and what, you know, our, what Trace three specializes in, and I think what we've really put a framework together on is like, overall AI strategy. What is your strategy? What are you trying, uh, to get done?

And, and it could be a couple use cases out of the gate, or it could be a broader strategy, um, around let's just say productivity internally. We gotta get, uh, we ship 10 pallets a day, we need to ship 20. How do we go work through that?

Right? Those types of things. What is the overall strategy?

I mentioned it before, and one, one of the things that Truth three specializes as well, is AI governance and risk. Are we having that conversation? Are we really working toward that AI operations?

What are some of the operations that are kind of in play along with the data quality aspects? So you start kind of carving some of that out. We, we ourselves inside of Trace three, we're like, there is no way we're gonna be able to help our customers, uh, if we do not understand it internally from our side.

And so we had kind of folks in that same mindset of like, Hey, let's get a couple of use cases, throw 'em over the fence, let's see how they stick. Now, copilot, Microsoft's, uh, copilot comes in and, and now everybody wants to go deploy that. Are you putting governance around it?

What's the overall strategy? Do you have an AI strategy? Do you have a center of excellence?

I know that's a bad term in certain organizations, other organizations embrace that. Um, but I would say that you're kind of seeing a lot of organizations kind of in the middle of what you were talking about where yep, we've got some use cases, we're gonna go work on that. Uh, but let's not get too far over our skis because we're not sure how big this actually can be.

Um, how do I work and balance that more in the middle? And that's kinda what we're helping and what we've tried to, I guess, form formalize and specialize, uh, on our side. Um, because again, I'll go back to it.

You can, you can have those three use cases, throw it over the fence. You don't have good data quality, you're not gonna get the results that you need. You're not gonna get good ROI, you're not gonna go see that, right?

So On the other extreme, you're starting to hear tales where people are creating these kind of like AI budgets that are like bills moving through Congress and all kinds of things are getting attached to it. And hopefully they're gonna get approved and it's gonna be like this kind of like major initiative where, you know, enables them to address all kinds of issues that are perhaps tangentially related to ai, but relevant to the project. Yeah.

Well, there's, and, and that's what we're seeing, Mike, right? Like, uh, we have one customer that has, you know, they called it out, they have X amount of dollars, um, from now until the end of the year, and they're gotta go do something, right? Um, what is that something?

Um, well, we need to kind of work through that and figure that out. Again, no strategy, no plan for growth after the fact. We just gotta go do something.

What is that something? And how do we go work through, through that? And, uh, what's ended up happening is, is that now that it's gotten out this particular business unit and this particular customer, now that that's gotten out, now their units are like, oh, they have funding, but we can just throw everything over the fence to them and then they'll figure it out and maybe we'll be the one that gets picked, um, that we get to go after, right?

And so it becomes like this aspect of, um, uh, jockeying for position where my use case is better than your use case, and we know that there's some funding, but let's go after that funding, uh, to be able to go do it. And I, you know, I don't see a lot of organizations, uh, going all in on like open source, uh, but we do see some organizations leaning on the open source act aspect of things. So, so that, Hey, you know what?

I, you, you talked about it earlier, build versus buy. I could go build it and then I'll deal with the results and the, and the funding and all this other stuff later on. Uh, but some of that is what we're seeing kind of real time, uh, from my perspective.

All right, folks, you heard it here. Well, you could argue that 2024 was the year of experimentation, but it's August already and 2025 is right around the corner. And some harder questions are definitely gonna be asked, Hey, John.

Well, I mean, again, you get, again, prioritize your data quality, focus on clear use cases, invest in your talent and, and training. You're gonna have to leverage robust structure. When I say ai, I don't want to be, I don't mean that as added infrastructure, because that's what we're seeing.

Hey, I need more GPUs, I need more stuff. Mm-Hmm. But you're gonna have to leverage that ose, you know, that really strong infrastructure to be able to go do that.

And then again, continually monitoring and optimizing. Again, it's all training. So get better, learn more, drive more initiatives, drive more use cases.

Now you're starting to go work through that. And I would also say probably that last thing to your point, um, start small, build gradual. If you're, again, if you're gonna do that on-prem, start small on-prem and, you know, be able to work it.

If you're gonna go pick a cloud, go pick a cloud and get good at it and, and, and lean on that particular cloud to go drive some of that. Um, and then lean on an organization like Trace three that can come and help and, and doesn't have the blinders on, can come help from an agnostic standpoint and help you go drive some of these initiatives. All right.

The only wrong thing you do is nothing. Right? Hey, it's great, Josh.

Thanks for being on the show again. Yeah, appreciate it, Mike. Thanks for having me.

Look forward to the next conversation. All right, back to you guys. Listen.