The Future of Industrialized AI Delivery

ModelOp CEO Dave Trier joined host Brandon Zemp on the BlockHash Podcast for a conversation about the widening gap between what enterprises are spending on AI and what they are actually getting back from it.

Enterprises have invested hundreds of millions in AI, Dave says, and most are still hitting singles. His diagnosis: AI work is fragmented across business units, each with its own tools, models, and vendors, and every new use case has to fight its way through legal, risk, security, and architecture one at a time. The result is what he calls the cottage industry of AI. Handcrafted, slow, inconsistent, and nearly impossible to govern at scale. His comparison is the automobile before the assembly line: high cost, low volume, quality problems, and compliance headaches.

The alternative is industrialized AI delivery, and Dave gets specific about what that requires, including the three things he would tell any Fortune 500 CIO to put in place first.

In this episode

• Why the enterprise AI bottleneck is delivery, not innovation, and how organizational silos plus layered approval processes compound into what Dave calls "enterprise frictions"
• The two mistakes he sees most often: teams that skip the process because nobody told them it existed, and leaders who over-trust AI in a technology that is inherently risk-bearing
• AI FinOps, defined: managing and optimizing AI value by weighing cost, benefit, and risk together, not cost alone
• Why token-based and per-API-call pricing turned AI cost into a CFO problem, and why cost has to be mapped to the individual business use case rather than to a model like GPT-5 or Claude
• The pendulum swing from "throw spaghetti at the wall" experimentation back toward disciplined, business-case-driven AI investment
• The three priorities Dave would hand a Fortune 500 CIO today: an AI system of record that treats AI as a first-class asset, security enforcement down to the network layer, and an industrialized AI delivery engine
• Why agentic AI was the tipping point that ended the Wild West era, because agents reason and act on behalf of humans
• How the CIO role shifts from managing software to operating AI ecosystems, becoming what Dave describes as the AI operating executive
• What is next for ModelOp, including the Kong partnership for top-to-bottom policy enforcement and agents built into the platform

Transcript

Brandon Zemp: What's up, guys? Welcome back to another episode of the BlockHash podcast, your number one resource for all that is happening across the emerging tech space. Be sure to like, comment, and subscribe wherever you listen to or watch the show. Today, I'm joined by another wonderful guest, Dave Trier, CEO of ModelOp.

Dave, welcome to the show. Pleasure to have you on. How's it going?

Dave Trier: Awesome. Thanks so much for having me. Looking forward to it.

Brandon Zemp: Likewise. Looking forward to discussing ModelOp with you today and chatting and kinda getting your take on a number of different topics.

To kick things off, maybe tell us a bit about yourself and your background, what originally sparked your interest kind of around AI, what eventually led to ModelOp. What's your story like?

Dave Trier: Sure. So I've actually been in the enterprise technology space for over twenty years. I've always been in this new emerging technology area.

So about five years before cloud was a big deal, I was working on that. And then, obviously, data science, big data came out. I was working on that well ahead of time.

And then this area around AI, governance, management, operations really became a huge wall that a lot of enterprises were hitting. So that obviously got my interest because at the end of the day, I'm about delivering, helping large enterprises deliver and use the transformational value of technology. So I said, there's gotta be a better way, like normal, right, of how do we help to overcome some of those challenges. So that's a bit of my background.

I came to ModelOp back in 2019. As I mentioned, I was routinely helping to deliver these large scale AI and ML projects and just kept hitting the same challenge around how do we go from idea to production very quickly? How do we overcome some of the typical enterprise frictions? And so I joined ModelOp along with some of our other colleagues that are here as well that helped to just turn this into a proper software company.

And that's what we've been focusing on for the past seven plus years.

Brandon Zemp: Exciting. And then for the listeners that maybe aren't as familiar with ModelOp, you know, for the first time, what does the company kind of do? What problems are you guys kind of aiming to solve when it comes to CIOs, CTOs, chief AI officers, and what they're struggling with today?

Dave Trier: At the end of the day, at the top level, all these enterprises have invested hundreds of millions into AI. And they start to wanna see the transformational promise of value that AI is supposed to drive. So we help large companies. We're a software company, but we help enterprises to industrialize AI delivery with the proper governance.

So think of it this way. I'll use an analogy. I love analogies. Right? So you think about these large enterprises.

I'm dumping hundreds of millions into it, and these teams are just continuing to use the baseball analogy. They're hitting singles. Right? They're kinda tired of the singles.

Right? They wanna see home runs, not just one or two home runs. They wanna see repeated home runs. So we, as ModelOp with our software, help to turn the typical cottage industry of singles that keep getting hit into something that is consistent, repeatable, automated.

Think of it as almost a factory like way to go from idea to actual usage of AI across the entire organization. That's what allows them to unlock that, again, transformational value of AI. We have customers that have publicly talked about, well, we want to generate one to two billion dollars of profit from AI. So you can't do that with hitting singles.

You gotta industrialize the overall AI process, and that's what we focus on day in, day out.

Brandon Zemp: Yeah. I love the baseball analogy, and it leads in so well to this next question too because I don't think that AI has an innovation problem. It definitely has more of a delivery problem.

Like, there's plenty of ideas. There's plenty of innovation and research and things being developed, but not actually being delivered or put out. Why is that such a challenge today for enterprises, you think? And what kind of goes into that?

Dave Trier: Yeah. Yeah. And you kind of touched on it already is that most enterprises, because they have so many different lines of business, business units, departments, teams, etcetera, it's very siloed. And each one of those departments and teams each have their own tools that they're using.

Some like to use open source, some cloud, some vendor. Right? And now there's a set of, you know, different vendor mature frontier models that they want to use, etcetera. So you've got a variety of different AI solutions that are being used for both development and running it.

You've got the silos organizationally, and then you've got complex processes that get layered on top from a legal data risk architecture security front. Right? And so you've got all of these different, what I call enterprise frictions, that slow things down. But at the end of the day, because of the silo nature, each team is doing bespoke development and trying to put out an AI solution one at a time.

It's slow. It's ad hoc. It's inconsistent. Again, to use another analogy, think about the car before the assembly line.

Each car was hand built. Right? It was slow. It was inefficient. There were quality issues.

Compliance was difficult. It leads to really high costs and low volume. And then the assembly line came around and said, okay. Well, let's turn this into a consistent automated approach to go from design to production rapidly, reliably, and ultimately, profitably.

We're in the same space as AI. Most of these teams are doing bespoke development, handcrafting them, artisanal type approaches to developing AI. That's inefficient. It's slow.

It's unreliable. It's hard to have compliance across that whole lot. So, again, that's some of the reasons that, obviously, we exist is to help them to overcome some of that cottage industry, we call it, and turn that into an industrialized approach overall.

Brandon Zemp: Yeah. So you think these friction points have, like, an adverse effect on being able to also, like, scale AI too.

Dave Trier: Absolutely. What happens, I've seen this time and time again, you've got some brilliant business and potentially AI data scientists, etcetera. They've got this great solution they developed.

They throw their hands up because they say, I'm done. I can't keep doing it. I can't keep hitting the wall. I keep getting blocked by these different teams.

I go to Jim in security, and he says, we'll talk to Jane in legal. And Jane says, okay. Wait. Now, because of this particular nature, you gotta go over here into architecture.

Right? So it's kind of a circuitous route that they have to take across all these different teams and processes. So, again, they just kinda throw their hands up and say, forget it. Done.

Brandon Zemp: Yeah. And these processes can be very complex. And sometimes it's cause and effect, A, causes B. But there's a lot of human element to it too. Are there mistakes that you think are being made at enterprise level that could be fixed and rectified that maybe clear a lot of this up?

Dave Trier: There are. And a lot of the mistakes actually are just unknown. Right? That you've got a business person or a team that has a great way to do something better with AI.

They just don't know the process, and so they just go ahead and do it. And so at that point, they have then skipped some of the checks and balances, some of the, again, good corporate guidelines, etcetera. So there are mistakes made on that front, just unbeknownst to them. They just didn't know what the process was.

And then there are also mistakes of just overly trusting AI. Right? At the end of the day, AI is inherently risk bearing. Right?

So there are mistakes to think that, well, AI will solve all my problems, first off, and then go forward and say, well, AI is always correct, which is not. So there are absolutely mistakes that happen as part of the overall process.

Brandon Zemp: What about spending? I imagine with generative AI, now agentic AI, the significant amount of infrastructure that needs to be built and what goes into that and the level of complexity there, it's extremely expensive for any enterprise to make that commitment, and many of them are.

Is spending something that also is a bit of a problem right now?

Dave Trier: It is. And it's actually become an even bigger problem more recently as you've probably seen that a lot of the AI companies out there have moved to a really a variable based pricing approach. So per token, obviously, is the most one talked about, but there's other ones around per API call and the like. So now you've got this place where you're kinda trusting that AI is gonna give you the value, but then also the cost is really variable.

And at any time, especially around the frontier models, that they can just change how they price their token usage. So it becomes a real challenge of how do I make sure that I'm managing the costs around those when there's not a known, you know, set amount that I'm being charged because it's that variable usage based approach. So it is absolutely top of mind, not only for CIOs, but CFOs. Right?

So how do you manage that problem overall? And at the end of day, what we talk about with our customers and we help through our software is, well, how do you manage and optimize those costs versus the value that we're getting? How do you weigh the trade offs between the costs, the risks, and the benefits? And so for us, it's about not only just making sure that we're collecting the costs and managing those, but also making sure they understand, well, how is that weighing against the overall value picture as well?

And specifically, the thing that I would say for listeners, the key takeaway is how do we map the cost back to a specific AI use case? A use case is the business scenario that you're using AI for. Right? How do you map the cost specifically back to a use case?

Because a lot of folks out there really think about, well, let me just get all the cost for GPT 5 versus, you know, Sonnet 4.7. It's not that helpful because you wanna actually tie back. Here's the cost specifically for this use case that is doing fraud detection within this business unit. I wanna know exactly how much that's costing because then I can do the exact rationalization.

Is that cost outweighing the benefits, if you will? So, again, real key takeaway in our focus is how do you map those costs down to the individual business unit, not only for chargeback, but that, again, cost benefit analysis that you wanna do.

Brandon Zemp: Do you guys have any use cases of some folks that you've worked with that maybe have seen some significant change in these different architectural elements for AI, whether it's spending or some of the other things that we talked about earlier where you've seen some change working with ModelOp, or even some use cases where maybe it's even like a full one hundred eighty too? I'm just kind of curious, like, who you've worked with. What's kind of stood out for you guys?

Dave Trier: Sure. So we've actually had a bit of our history in the regulated industry. So we've done a lot of work with financial services, insurance, as well as pharma, health care, etcetera. So that's a bit of our heritage.

And I guess just a couple things that we've seen change over time is early on, there was a big push for just go and experiment. Just go and experiment, throw spaghetti at a wall, and we'll see what sticks. Right? And now more recently, especially given our conversation just now around cost, they've thought about and said, we need to take a more disciplined approach around how we do this.

Right? We don't wanna just throw everything at AI because it can really drive token usage up, which obviously turns into cost. So I think more recently, we've seen that shift. I wouldn't say it's a full one eighty, but it's gone back on the pendulum.

Right? It's swung back on the pendulum towards a more disciplined approach. Let's make sure we lay out, hey. Here's what the, obviously, the scenario or use case we're trying to solve, but also here's what we believe the business impact is going to be, both from a process perspective of people and also, obviously, that financial perspective overall.

So, again, just one example of a bit of the pendulum swing, if you will, in this space.

Brandon Zemp: There's a topic gaining a lot of attention that I've heard of called AI FinOps. Is that something that is relevant for you guys as well? Do you wanna define that a little bit too?

Dave Trier: Sure. Sure. So at the end of the day, ModelOp, we obviously have our core software, but that's composed of a couple of key elements. The first is an enterprise AI system of record. So you wanna know everything about AI across all the different teams, departments, technologies, you come to ModelOp. Part of that equation is the FinOps component so that we can understand, again, everything about AI that includes the financial portions of that.

Now in terms of definition, for me, I would say FinOps is about how do you manage and optimize AI value, which, again, comes back to weighing the trade off between the costs, benefits, and risks. People often don't, you know, remember the risk part of the overall AI value equation, but it's not just cost and benefit financially, but it's also the risk. Because even if there is a higher benefit over the cost, if the risks are pretty high, that kind of delta isn't really worth it to a lot of organizations. They don't wanna have the risk, whether it be regulatory, brand exposure, obviously, financial as well. So sometimes, again, that delta, even if it's higher benefit than cost, the risk will outweigh it. So, again, for us, it's about managing the cost benefit and risk in that equation.

Brandon Zemp: If you were advising a Fortune five hundred style company today, what are the three biggest things you'd prioritize or prepare for a CIO or someone in an executive position to prepare for enterprise scale level AI today? Like, if you could give three things, what would it be?

Dave Trier: Yeah. I think the first would be what I mentioned is the AI system of record. Right? AI has had so much investment.

There's so much pent up value that's behind it that it needs to be managed as a first class asset, just like a customer record would be in the CRM, just like a, you know, IT service record into an ITSM system. So first and foremost is having that AI system of record. So you know everywhere that AI is being used, and you know everything about it all the way down to the risk and the finances. So that's the first one.

The second one, I would say, has to be around security, right, so that you make sure that you are having down to the network level understanding and policies in place to contain when AI goes wrong because it will go wrong. Obviously, you've seen recent headlines about it with some of the breaches. I won't name names, but you've seen that in the headlines. So that's number two.

And then the third thing I would say is making sure that you have an industrialized approach to delivering AI, especially for the CIO. That's kind of their main job, right, of how do we make sure that we have an enterprise grade, scalable, consistent, financially optimized approach to using technology. So that's where the third component would be that AI delivery engine that helps to, again, take — like before the Model T — the manual approach of developing a car into something that is repeatable, the factory type model.

So those would be the three that I'd recommend.

Brandon Zemp: What do you think enterprise AI will look like over the next three years down the line?

Dave Trier: Yeah. That's a good one. I think if you just take a look back at some of the history of technology, the biggest winners really weren't the ones that went and bought the fastest, either database or cloud instance or, you know, CPU, whatever it may be. It was actually the ones that had really thought about the strategy about how do I turn that, like I said, artisanal approach into something that's industrialized.

Because, again, like, we go back to that baseball analogy. It's great if you get a single or double. That's great with the fastest cloud instance or database. But to do so repeatedly, that's where you start to stack the wins on top of wins on top of wins.

Right? And so that industrialized approach around how do I go and consistently take that AI that has a great idea and turn it into, you know, production with the right standards and governance and security and operations and really operating models. So that, for me, is the first piece that I would see. The second thing that I would say is that more of turning from what IT is today as managing pure software.

Right? So just managing the billing system and the like. And now it's about running ecosystems, AI ecosystems. Right?

Because these AI ecosystems will involve a variety of different agents, multiple LLMs, tools. They're all talking together, if you will. So, again, IT shifts from just managing software to managing these AI ecosystems. And above all that, right, the CIO, again, is more than just managing these software systems, but it's about being that AI operating executive.

Right? They're the ones in charge of operating those ecosystems, making sure that it's industrialized, well governed, and then again, financially beneficial to the entire enterprise. So just a couple thoughts.

Brandon Zemp: Do you think most companies will need a chief AI officer? I feel like that would be a more common position now going forward.

Dave Trier: So I do see that coming, at many of the enterprises we look at around the chief AI officer. I think, at the end of the day, as long as there's a tight relationship between a chief AI officer and a CIO that is able to turn that into, like I said, that operating paradigm. Right? The AI operating paradigm. I think that as long as you have that tight relationship, that works. Some companies aren't probably big enough to have both.

But the largest ones, yeah, you want a real focus on AI, the AI architecture, etcetera. Yeah, a CAIO makes sense.

Brandon Zemp: There's a lot of talk about the AI space being in a bit of a bubble with how fast it's moved over the last three years or so. I think ChatGPT came out in, like, twenty twenty three. So, yeah, like, three years. It's advanced super, super fast. Some people think maybe it's putting one foot in front of the other a little too quickly.

Some of the things we talked about too here today in the episode, I think kind of allude to that with deployment and pricing and what's being done in terms of delivering product versus innovation. Do you think that it's bubbled up a little bit too much? Or do you think that we're just kind of getting started? I mean, I don't really know if there's a right or wrong answer to this, but there's so much excitement and not enough talk, I think, about like where it's at as such a young space.

Dave Trier: I mean, with any new technologies, there's always the ups and downs. Right? Obviously, we're in a very high hype cycle as you can imagine right now. So, of course, there's gonna be peaks and valleys.

You see that every time. I do think that this is such a fundamentally transformational technology that it will be there to stay. It's not one that's gonna die, if you will, like some of the prior technology investments for sure. But, yeah, I would say it is a bit of a hype.

But, again, that's more of just the capability of what it can do. I think it — like I said, it behooves — again, I'm talking about enterprise here — the enterprise to be able to corral it, right, to rein in the Wild West of AI, to turn it from, like I said, just having the one off pockets of wins into something that is just part of how they do business. So that's the unknown.

That's the thing that we'll keep watching. Obviously, we've made a big bet that, from our company perspective, that software like ours is able to help them to shift from that artisanal approach to something that is just part and parcel of how they do business. But we'll see. I think that's an area that, you know, obviously, is something that most enterprises are trying to do right now.

Brandon Zemp: Are there any trends that you see that companies should be paying attention to? It could be trends in terms of, like, agentic and AGI and generative AI and all these different, like, subsets of where AI is going, or it could be other intersections with other technologies. It could be AI and IoT. It could be AI and blockchain. It could be AI and specific industries.

Anything that stands out for you?

Dave Trier: I think, I mean, everybody would say this, but it's so true. It's the agentic side of things. Right? So we've been doing this for many years, and what we've found is the tipping point has been agentic AI where especially the CIOs and other executives just said, okay.

Enough's enough, guys. Like, we let you guys have a bit of the Wild West of AI for a while, but agentic has been that tipping point. And that's just because they are acting on behalf of humans. Right?

They're doing the reasoning. Right? And then they are making actual decisions around those. That's the reason that I see that's been the tipping point, you know, just in general across the whole AI realm.

So, yeah, agents has been a real — it's been fun with the agents world coming, agent chaos, I should say, coming into play.

Brandon Zemp: Speaking of agents, there's a lot of people that are theorizing that we're gonna see the first billion dollar company ran by one human relatively soon, thanks to the power of what agents might be able to do. There's some companies that are also kinda like pushing that along a little bit, making it a little bit more of a possibility. Do you think that will actually happen, or is that just a lot of marketing and fluff?

Dave Trier: I think that might be a little over the top. The billion dollar company. Because at end of the day, the biggest unknown is just how do you make sure that you ensure agents are operating within the confines that you put forward. Confines being from a business perspective, compliance risk, security, etcetera.

We just saw this past week, you know, one that got out of its confines. Right? Even with one of the biggest players in the space with all kinds of security and sandboxes in place, and it got out. Right?

So I think that's a bit over the top around it. But you definitely do see organizations, and even large enterprises, taking a different approach of saying, okay. When we're doing an enhanced business process or a new business process, can we make it AI first? Can we do it with AI first?

So that shift is definitely happening and will continue to happen, of course. We do the same internally with our own company right now. It's just, alright, when we're thinking about new capabilities, new extensions, even new go to market models, etcetera, we first ask ourselves, okay. Should we build this from the ground up with AI if it's possible?

And if so, what kind of efficiencies can we gain around it? So I do see that shift. But one person billion dollar company, I think we're a long way from that.

Brandon Zemp: So what's on the horizon for ModelOp the rest of this year? Any important updates that we should be aware of? Any exciting partnerships? What's on the horizon?

Dave Trier: Yeah. Absolutely. So we actually just announced about a week or so ago, exciting partnership with Kong, who handles the API gateway layer. Right?

So they're a security vendor that handle that network level enforcement. So we have a very nice and tight partnership with them to be able to do top to bottom enforcement of the different policies. So that's an exciting one. Overall, we've been doubling down on our partnerships.

We'll continue to see a number of new announcements coming out over the next couple weeks, so that's been a heavy investment. Of course, in parallel, we have over the past six months been leveraging agents ourselves. Right? So leveraging agents within our product, so offering our customers to allow them to streamline that industrialized process even more with agents and to bring their own agents so that they can say, okay.

Well, I developed this agent which helps me with this very specific task that I have to do that's part of my company. So they're able to plug those into ModelOp and, again, just help to further streamline, further make that process more efficient.

So there's a couple other announcements, but, yeah, partnerships, leveraging agents absolutely been front and center for us.

Brandon Zemp: Excellent. And then if anyone wants to learn about ModelOp, if there's maybe a company that wants to become a client at some point, someone wants to get more information, where should they go?

Dave Trier: Yeah. ModelOp.com is the place to go. I'll leave my details here. You can reach out to me directly. Love to have the conversation. But ModelOp.com, can find all about what we do, our customers, our approach, and happy to chat anytime.

Brandon Zemp: Awesome. If anyone wants to connect with you as well, do you have a LinkedIn or an X, or are you present on socials?

Dave Trier: I do. Yes. I'll leave my LinkedIn that you can send out to your listeners. And, again, please DM me at any time. I love talking about this stuff.

Brandon Zemp: Awesome. Yeah. We'll put everything in the description for the episode below so everyone can find everything super easily. I appreciate it, Dave. It's been a really good episode learning about what you guys are doing at ModelOp and talking about the industry and enterprise and where AI is kind of intersecting these days, some of the problems there, some of the things that need to change.

It's a very exciting time. So keep up all the great work, and I would love to do this again too. Really good conversation. Appreciate it.

Dave Trier: Awesome. Thanks again for having me. Have a good one.