AI Agent Builder Summit | AnalystANGLE Wrap

Clips
News
More from AI Agent Builder Summit

Scott Hebner

Principal Analyst, AI

theCUBE Research

Dave Vellante

Co-Founder & Co-CEO

SiliconANGLE Media, Inc.

Race to agentic AI accelerates amid rising stakes for startups and enterprises: theCUBE analysis

The convergence of cloud-native infrastructure, harmonized data layers and real-time automation is opening the door to agentic AI — intelligent systems that act autonomously and collaboratively across complex workflows. This evolution is not just technological but deeply cultural, requiring enterprises to rethink both operations and organizational models to keep pace with rapid innovation.The industry is undergoing a major inflection point as traditional automation gives way to dynamic, situational intelligence powered by artificial intelligence agents

play_circle_outline Introduction to AI Agent Builder Summit and analysis of key takeaways

play_circle_outline Discussion on the transition from generative AI to agentic AI

play_circle_outline Significance of infrastructure in enabling agent layer functionality

play_circle_outline Startups' advantage in the agentic AI market

play_circle_outline Technological roadmap for agentic automation

play_circle_outline Evolution of business logic from hard-coded to dynamically created by agents

play_circle_outline Empowering Agentic Systems with Generative AI, Explainable Decisions, and Real-Time Deployment Strategies

Info
Transcript

AnalystANGLE Wrap

Scott Hebner

Principal Analyst, AI theCUBE Research

Dave Vellante

Co-Founder & Co-CEO SiliconANGLE Media, Inc.

HOST

Scott Hebner, John Furrier and Dave Vellante of theCUBE lead a dynamic analysis session from the AI Agent Builder Summit, unpacking the top insights and industry signals from this inaugural event. With 20 speakers across 12 organizations including Semaphore.ai, Deloitte and IBM, the conversation reflects on how agentic AI is reshaping business models, tech stacks and enterprise culture.

Drawing on decades of high-tech coverage and 10,000-plus executive interviews, the panel discusses trends in AI infrastructure, the growing role of semantic layers an... Read more

explore Keep Exploring

What were the key takeaways from the AI Agent Builder Summit? add

What is the key factor in the transition of generative AI to agentic AI? add

What is the relationship between predictive analytics, GenAI, and infrastructure transformation in companies like Nvidia and IBM? add

What factors are shaping the agent market and potentially favoring big incumbents, according to the speaker? add

What will be key in the technology area roadmap to enable vertical and various types of agents, as well as the acceleration of infrastructure development? add

What is the timeline and process for the evolution of digital assistants and intelligent agents in the future? add

What are the steps involved in building and deploying real-time agentic systems using generative AI and LLM-based infrastructure? add

bolt Powered by CUBE AI

AnalystANGLE Wrap

search

Scott Hebner

>> Hello. Thank you for tuning in. Welcome to the AI Agent Builder Summit. This is the Analyst Post Game Show, where we'll wrap up today's summit and analyze what we've learned during the summit. We'll help you separate the signals from the noise. I'm Scott Hebner, the principal analyst for AI at SiliconANGLE Media and theCUBE Research, and thank you for joining us. The AI Agent Builder Summit, which is the first in the series of agentic AI summits that we will broadcast this year that are dedicated to exploring the transformative value of agentic AI, and more importantly, to help you gain insights and deploy best practices by hearing from industry pioneers with real-world experiences. During the summit, you'll hear from 20 of these industry pioneers representing 12 companies, including software and platform vendors, Semaphore.ai, Ascendion, Agilpoint, Deloitte, SAS, IBM, and Geminos.ai. You'll hear from a strategic vendor independent point of view by Deloitte, and the perspectives from those actually on the ground, end user businesses, Fitch Group, Emerson, Align Technology, HP and G2. We've learned a ton, and we thank all of them for making this an invaluable experience for those of you that are just getting started on your journeys to agentic AI. To help put all the pieces together and complement what you've heard with their extensive experience in the enterprise tech marketplace, please welcome the co-founders and CEOs of SiliconANGLE Media and theCUBE, who have collectively interviewed over 10,000 leaders in enterprise high-tech over the years, Dave Vellante and John Furrier. Guys, welcome to the show. Thanks for being here.

>> Great to be here. Great event. Phenomenal. Thanks for having us.

Dave Vellante

>> Hey, Scott. Hey, John.

Scott Hebner

>> How are you guys doing? Yeah, it's been been a wonderful experience. We got to hear from a lot of people. 20 people is quite a bit in one day. We filled the whole day, and what's interesting about each of the sessions I found was that there was a little bit of different perspective and aspect of agentic AI. It wasn't a lot of redundancy, so I think if you go through all 12 of them, there's a lot to be learned there. So let me just start just quickly from what you saw, what are your key takeaways? John, maybe we'll start with you.

>> Yeah. My big takeaway was one, first of all, the timing of this event is really perfect because this agent-builder focus is the key because people are building out. Now, as the AI infrastructure, as we've been reporting on SiliconANGLE, theCUBE and theCUBE Research, AI infrastructure is advancing very, very fast. Even with the tariffs and all that nonsense, the speed of advancements on the horsepower is enabling this data layer that we've been following some time. Some call it a semantic layer. Dave Vellante, George Gilbert have been documenting aggressively this harmonization piece. This is all setting the table because platform engineering is merging in with AI. So this AI agentic layer, I call it the middleware for the lack of a better description, is exploding with innovation, both from a business transformation and technical transformation. We're going to see conversations on theCUBE here, the business and the action for the people on the ground and the trenches, but also there's an alpha technical thing going on as well, because without data, there is no AI, and that is well known and talked about. So to me, I think that the market is absolutely transitioning very fast. It is transformative because of the things like digital twins. Like the business value, not just the cost efficiency and bottom line benefits, there are real revenue impacts. So all theaters are exploding in value and it's happening both in the startup ecosystem and the big company ecosystems, and that is really fantastic. So on the playing field of AI, this is a super active and exciting market, and people will win and some people will lose, so we are watching it very closely. So very strong program and I'm looking forward to continuing to do more conversations on this.

Scott Hebner

>> I did find it interesting, a lot of the speakers talked about the word coworker. Not so much AI agents, what they're really trying to do is create coworkers that partner with humans to get things done, either individually or as part of a workflow of a process. And so yeah, I really think this is building off that semantic layer and taking it to the next level. Dave, what about you? What were your key takeaways?

Dave Vellante

>> Well, I wanted to address the why everybody's so excited about this, and I wanted to start with the macro and this whole thing, I often use the term super cycle, the data center super cycle. And people don't like that term, but John, I love it. Super cloud, super cycle.

Scott Hebner

>> Super chips.

Dave Vellante

>> Super chips. So I wanted to help you understand how big this trend is. Andrew, if you'd bring up this slide. So if you look, this is spending on data, the data center build out. So it's space, power, cooling, compute, storage, networking, the core infrastructure. And you can see for years the brown, that's the traditional workloads. The blue is AI workloads, but for years, data center, the whole bar has hovered around $200 billion. And then it jumped from 220 billion in 2023 to 350 billion last year, and it's going to hit a trillion dollars early next decade. So it's like a 25%, 10 year CAGR. It's massive. Let that sink in for a minute. In one year, the growth went from four or 5% in data center to almost 60%. One year. Why? Well, of course it's because AI infrastructure, and now the AI piece of that is dominant. And John, to your point, so if you see here, so we take that previous chart and then we isolate on the AI piece of it. The dark blue is the cloud and the light blue is the enterprise, and so to John's point, the value today is in consumer. It's Meta's targeting ads, it's Google with Search, it's ByteDance, TikTok and other social media companies. That's where the value is today and that's where the spending is today. The vast majority of the spending, like 95% of it, is from the hyperscalers. And most of that is, again, serving the consumer, ChatGPT, Perplexity, et cetera. And this is important because consumer markets are where markets are formed. It's the value, the volume, economics, and of course the speed of innovation. You call that the third V, the velocity. But the enterprise doesn't hit the steep part of the S-curve like cloud did last year or AI did last year. The enterprise, doesn't hit until '27, '28 timeframe, so we're still... Thank you, Andrew. We're still in the formative stages of this AI, and so when everybody gets excited about agentic, there's a lot of work that has to be done. And my takeaway, Scott, to answer your question, is I think that came through. When you take the vendor hype and you map it with the practitioner realities, there's a journey that has to take place, and you and George talked about the four ladders. So a lot of work has to get done. It's not going to happen overnight and there's going to be incremental value all the way, and then once it hits that steep part of the S-curve, people are going to be printing money and that's when the big disruption comes.

>> And just to add to that, Dave. You point out the growth, which by the way, it's phenomenal research. Great job there and good work. I would also add that it's compounded on scope from a magnitude standpoint, is it's not just the enterprise, it's the startups. And a lot of people are trying to understand, is this a startup game? Is it the big guys in the cloud win? We heard Jensen Huang say most of the inference is happening in the cloud, so yeah, the cloud folks will win. Google Cloud was talking about native AI making their products better, but their customers are not only consumers, they're the startups as well, so both the enterprise and the startups are winning. So I would say the way to scope this out and think about the frame is think about the SaaS market with the cloud. Dropbox, Airbnb, they were hitting white spaces that were too small for the big companies because why staff up all that resource? And that was a different market. You build an app, you put it in the Apple Store, people download it, so it was much different. The vertical agent market where you have domain expertise could be almost two to three times bigger on the startup side than SaaS. So SaaS is moving to agents, so the metaphor there is on the startup side, you're seeing the cycle of startups, and I love the term super cycle because you're starting to see the innovation, because if you have domain expertise, you can get into the agent. But also the playing field's level because the cloud, the banks, the companies in the enterprise also have agility, so they have domain expertise and they have the data. So what this means is it's a speed game and whoever gets that value first will win. So both theaters of startup ecosystem and enterprises are exploding in value, and I think if we look at the SaaS market in comparison on the startup side and what that meant for businesses, it wasn't a one-to-one speed game. It was, yeah, banks got a nap, but startups became unicorns and decacorns. So that's different with agents. It's all happening at the same time, so I think you're seeing new logos pop up, you're going to start to see these white spaces develop, and this SaaS conversion to agentic is happening. And I think, yeah, a lot of work to be done, but if you're an investor and you want to build a venture, if you're an entrepreneur, the game is on. So that to me, Dave, highlights the customers. On the agent side is the enterprise. So all that together makes it even more powerful, and that cuts through the hype. Now the hype is legit, but who will win will be who can move faster. Look at AWS, moving at lightning speed. Google literally in the past 12 to 14 months, I won't say pivoted, just accelerated. And so the clouds are scaling and the startups are scaling, so again, everyone wins.

Scott Hebner

>> And there's no doubt that the spending in the marketplace really is starting to accelerate here, but let me step back here and just think of the big Uber transformation over the last couple of decades, probably 20 years or so, as we went from interconnecting everything to instrumenting everything to changing the economics of everything with cloud. Now we're into the intelligence phase. People were deploying predictive models up to a couple of years ago, and then we got into this generative AI phase which was the rage of everything. That was the hype at the time. That was only three years ago, and now we're moving on to agentic AI. And some of the conversation that we've had during this summit was, is this hype again? Is this a progression or is this the truly transformative part of the cycle? And it's like these presidential elections, everyone is the most important one of your lifetime. What do you really think is going on here with the move from generative AI to agentic? And Dave, I'll start with you on that one.

Dave Vellante

>> Well, I think all this talk about agentic transforming organizations, well, it's super exciting. As I said before, it's going to take some time. There's a lot of work that has to be done before the broad market can really transform into this agentic, autonomous enterprise. There's a lot of early examples. Microsoft with Copilots, Benioff calls it Clippy. Salesforce with Agent Force. Nadella says, "Well, the future software is our agents, agents like ours talking to backend crud databases," depositioning Benioff's SaaS business. But interesting comments, it didn't get lost on me, John. You said that this new wave could be bigger than the entire SaaS business, but the implication that Nadella is saying is SaaS gets commoditized, but you're seeing companies like ServiceNow and Palantir and Workday, Oracle, Amazon, Blue Yonder, they're all investing, pushing agents. Google with Agent Space, we saw that last week. So lot's going on, but as I said before, the vendor marketing is way ahead of the value realization in the enterprise, probably by 18 to 24 months but it's a genuine opportunity. I'm going to stop there. I'd love to if we have time, Scott, share the roadmap that George and I worked on with you. But I'll stop there and see if John has any thoughts on that comment.

>> I think the generative AI transition to agentic really is about scale. Genarative AI ushered in the era of, "Oh my god, that's magic. That's cool." And if you look at all the applications that are going into production, I call it where the rubber meets the road, the big discussion is how do you move from demo to scale? And demo, I mean it's working but it's not working at scale. And if you look at Nvidia, you look at IBM, they all have the same graph. Predictive analytics intersects with GenAI and the centerpiece is the new infrastructure. And if you look at the squint through the trends and look at Nvidia, all the change in transformative action is around memory, storage and networking. And we heard that clearly at Google Next .we're going to hear it from Amazon in June when we do our Amazon Ecosystem Leader Summit on June 19th. You're going to see the same thing and it's all the same, that the infrastructure is transforming, and that will dictate to me the agentic layer as an infrastructure piece, which will enable the GenAI experiences that will be rendered in the apps, whether it's vertical agents or agents talking to other agents. And that's going to be a lot of plumbing, a lot of data management change, a lot of memory architecture and how storage is provisioned. Hammerspace just scored a hundred million dollars, looking at what they're doing. Everything is being retooled for AI and that is transformative, and that's going to probably take about a year or so, as Dave pointed out, but that is real and that is actually happening. So you've got to look to the infrastructure pieces. That will tell you what's going to happen at the agent layer, and that's infrastructure. If it doesn't scale, it doesn't work, and that's where the agents will either thrive and survive or crash and burn.

Scott Hebner

>> Yeah. My point of view on this is GenAI to me is the equivalent of the browsers were in the early internet age. It was the most strategic decision you can make, but what happened over time is it faded into the infrastructure and it at best became a preference. And the value the enterprises got were how they built on top of and around that gateway into the internet called the browser, and that's where the value came from. I think with GenAI, you've got a very similar phenomenon here which is it is the gateway into the world of AI. It's democratizing everyone to use it. It's the enterprises that start building around it and on top of it with the SaaS apps and with specific models to their business, and that is in the agents of course, and I think that's what we start to see happening here as time goes by. So let's do this. I got one question for you guys that I want to ask, and then I want to go into success factors that we talked about during the summit, and then Dave, we'll end on talking about the roadmap. So let me ask the question I had in mind here, which is you guys have been doing a ton of work with the New York Stock Exchange, with the Wired program. You're out here, John, in Silicon Valley. Is this a great time to be a startup?

>> I think it is. People were pooh-poohing agents as it's going to be the big... Guys are going to win. Like I said earlier, everyone leveled up at the same pace in terms of capabilities. I think that the agent market is going to be very vertically focused because the domain expertise and the workflows, we've been covering workflows and data as the key value, I think that favors the big incumbents, unlike SaaS, which as I said earlier, was a white space, so I think that's one advantage. But at the same time, new innovations, if done faster on the startup, you're starting to see startups use AI better and faster than the incumbents. So I think if you're a startup and you're not playing the speed game, you're going to be toasted on this thing. And what I mean by that is that in the startup world, and before with SaaS, you've got success in the app store, on the cloud. You double down, you add staff. AI goes the other way. You actually reduce your staff because agents will do the work. You mentioned coworker. The Copilot model of the benefit favors the startup because now the startup can do more with 20 people. And whether it's a go to market or managing workflows or whatever it is, both technically and in business, where in the past, you would have to staff that up. So that was a little bit of a build out mode from a startup standpoint. That now has changed, so that favors the startups. And the startups are working at scale, and a lot of the hot startups that are out there, the young guns and the senior folks that know the systems, they're taking advantage of the infrastructure. So all the action right now is at the infrastructure layer that's between the agentic data layer. That's where the action is, and to me, if you're a startup, this is a great opportunity to win, but you've got to move fast,

Scott Hebner

>> Dave?

Dave Vellante

>> I definitely would agree with that. I think the last decade for startups, was defined by cloud native, or those companies that maybe didn't start out cloud native but very rapidly pivoted to cloud native. And you think Snowflake, you think Databricks. Even companies like Cohesity which started really on-prem, then brought their stack to the cloud. And so now, today, it's all about AI native. Why is that important? It's important because startups today can do 10 times the software development output than they could 10 years ago, so it costs much less to get a product to market. To build an MVP, to get a great product to market, to get product market fit, you can compress that timeframe and then get to product led growth much, much faster. I think you're still going to need a lot of money to do promotion and go to market. I think that will become more productive, but I think it's almost in some ways a zero-sum game because everybody's going to get there fast, so you're going to be competing very aggressively against others who are doing the same thing. But I think it's a fabulous time to be a startup. I love startups at the beginning of these waves or in the early part of these waves, maybe a company like OpenAI who started the wave. We'll see if they make it through, but now we're three years in and I think it's a great time to start a company because you can capture the signals and really place your bets in a big way.

>> And I think I would just add to that, David. I'd say that our observation on theCUBE is if you don't have the data, you cannot be successful. So in the enterprise, if you're a startup, you've got to look at the market and say, "Okay, is my incumbent that I want to take down..." Because startups are punching up. They always punch up, which we love. And if they assess the fact that they could move faster against an incumbent and win, that to me is the opportunity, and I think that's going to come back down to the platform engineering. And remember COVID, when we talked about who was winning in COVID? It was the folks that were already in the cloud. So I think if I was a startup, I would look at who's in AI as an incumbent, and if you're going to take territory in a big market or a slice of a vertical, which could be very lucrative, you've got to look at is the competitor more AI native than me? And if that's the case in your favor as a startup, you have a chance. And I think just like COVID, if you weren't in the cloud, you were screwed, and so the folks that were in the cloud took advantage of that because everyone was working at home. Same with AI. If you are in AI, you have to stay in it, you have to constantly be moving the needle. And if you have a competitor that's big and has the data and they aren't moving fast, I wouldn't invest. It would be a tough call.

Scott Hebner

>> Now, one of the areas I've been doing some research on is the emergence of these AI agent marketplaces, and I came across one startup that's I think a year and a half old, is very young. HOAi, Home Owner Association AI, and they actually started from the ground up and built AI agents to help homeowner associations do the things that they do. And when I was talking to them, they've had hyper growth and they tap into these agent marketplaces where people buy them, customize them and integrate them in. And that's consistent with what I've read like on McKenzie for example, that majority of businesses are going to buy pre-built agents, customize them and integrate them with the proprietary agents. So there's going to be a whole marketplace that's going to develop here, and I think that makes this a really interesting time to, as you said, to be a startup and build on gen AI and add that value on top.

>> Well, it's interesting, in the old days of startups, the question I would get a lot is where's the white space? Because they wanted to get into a white space and then sell to the big player, and if they get lucky, they can maybe sequence to a broader market opportunity. That was the classic playbook. Today, a white space could be an entire category, like homeowners. You could run the table on a category as a startup if you get it right. And again, to Dave's point, the ability to get product market fit is so much faster. The leverage you have with AI from a workforce and cost of capital perspective is so much lower that you're going to start to see lower funding rounds and category dominance in what I call these vertical agents. So I think that's the new SaaS. Who would have thought that homeowners associations? You look at the TAM, it looks small on paper if you were in the old world, but in the new world, it's a viable opportunity. So I think the vertical domain-specific areas, you can build a feature with agents, take over a category. That would be my investment thesis as an entrepreneur and as an investor.

Scott Hebner

>> I think one of the areas to watch on this too over time is if the analogy that I made to the internet, the early internet age with browsers is right. At some point, they had to standardize so the plugins can go everywhere, and we have to wonder if that's going to start to need to happen with GenAI if people are going to build on it and you can buy these agents and deploy them everywhere, so we'll have to see where that all heads. Let's make a transition here. I want to talk a little bit about what we think we heard from everybody in terms of success factors, and let me bring up a chart here. We talked a little bit about this on the pregame show. These are the five key success factors I think we heard about. First of all, the game is played in the field, not the dugout, meaning you need to get started and at least start experimenting and building the skills and the technology base for agentic AI, because these innovation cycles, they're moving at warp speed now. If you fall behind, you may never catch up. So get out there, start putting some points on the board. Simple, straightforward use cases, build from there, and start to assemble things on. On the bottom layer is keeping in mind that with this, if it's about coworkers and it's about making decisions and problem solving and achieving goals, it's about your organization and your people. It's an HR game too, so culture is not just part of the game. As Lou Gerstner famously said, it is the game. So spend a lot of time on your organizational culture about the agents are not here to take your job. They're here to give you superpowers and they're going to help you and us as an organization do more with less, and that's a good thing. Don't forget about the culture. Then within that, I think what was said very clearly is you got to have really crystal clear ROI objectives. Codify it, know what you're trying to achieve before you get started, and do them in increments. Two is in the end, it's all about trust. It's the currency of innovation now. No trust, no ROI. If you're going to have agents help you make decisions and problem-solve, they have to be trustworthy, because you're not going to just blindly trust an agent telling you to do something unless it has a good explanation why. And then finally, it's about empowering the talent, this notion of superpowers and coworkers, and I think we got to be very careful with this idea that it's going to replace everyone. It's definitely going to lower cost for an organization, it's going to allow you to do more at the same price point, but it's about your talent too. And so from a business perspective, this is what came out as we went through all 12 of these sessions, and I'd love to get your thoughts on this. Dave, maybe we start with you on that one.

Dave Vellante

>> Yeah, very interesting. I can't see the chart so I was just taking notes. You've got to be in the arena, you got to focus on organization and people, build trust and do so with clear ROI objectives, and you got to empower your talent. I would say those all make sense to me, and I guess I would add, you got to get your data act together and I'll talk about that in a moment, and I'm sure that was threaded throughout many of the conversations that you had. The culture thing is interesting, Scott, to me, and it comes back to data. I think that organizations are still, in my view anyway, not exceedingly data literate. You'll hear companies all the time talking about, "We're data-driven. What are the metrics?" But the key is, okay, do you have the right data? Is that data harmonized, and do you have the right metrics? Are organizations working toward those same goals? And I would say in many cases, because data is so stove piped, that oftentimes the organization's goals are stove piped, and so that creates friction from a cultural and people standpoint. So I think those are really good key principles that you've laid out. It's not trivial to get them all aligned.

>> I think the culture thing is huge. Andy Grove and Gordon Moore at Intel back on the last super cycle of the PC revolution, which had hardware and operating systems and applications, a similar three stage stack we're seeing now, full stack as they call it. He said, "We're Moore's Law," and the cadence of Moore's law was a cultural issue. And I think culture is huge, I think, and with agentic and what's going on with AI right now, it is a transformative transformation at a technical level and a business level. So business model, business transformation is happening on top of digital transformation, and to me, that will change culture. So I mentioned speed. If you're not at the cadence of speed, culture is the game. It's the only game in town because if you don't have the speed, you don't win. If you're not in the game, on the field with AI, you're going to lose to somebody else who's going to be faster. So I think like Moore's law, the cadence of Moore's Law and the famous Intel DNA when they were in their heyday will apply to the culture today. And you look at the startups and the companies and the enterprises that are being successful, even Dell, who we've been covering like a blanket, they're moving fast. And Michael Dell knows super cycles when he sees them and he's transforming Dell, and the startups also recognize that speed's the game. So I think culture is absolutely fundamental. It's one of those touchy-feely subjects. "Well, we have to need a good culture." No, no, it's a systems culture, it's a speed culture and it's a coding culture. If you're not coding with code assistants, using AI to help code move fast, not transforming the business model and how you make money, you won't be successful, and I think this is a key piece of the puzzle. So great insight, and that's why we have logos we're seeing on the screen in this session, they're not just the big names. You have upstarts, you have incumbents. Again, everyone's on the field. I love that dugout because also not in the seats either. Forget the press. The press just can't even keep up at this point, so I think the game is definitely being played on the field or the arena and it's going to be a bloodbath, but there'll be winners and it'll be big winners, and sometimes I think you're going to see some names we might not recognize. And so again, all super cycles, new brands emerge, they come out of nowhere, and like SaaS, we saw that play out beautifully. Who would've thought Airbnb would've been a big company? So I think you'll see agents make that same impact.

Scott Hebner

>> And just as an example of what Dave said about the data layer and the semantic layer. AgilePoint at one of the sessions and one of our sponsors, they actually have this layer that they called holistic abstraction, which Dave, as you and George Gilbert define it, it sort of does that, right? You've got to abstract all that data and all the different applications and make it into a semantic layer. And they've been around for some time, but they're in a scale part of their business, and you're right, this can really catapult a lot of businesses. The other thing before we move on to the technology roadmap, I just want to get some thoughts on that before we wrap up here, is Abdi Goodarzi, who is the head of generative AI at Deloitte, he made a really interesting point, John, to your point there on culture. He said he envisions and he's starting to see clients treat their chief AI officer as part of the HR function of a business, because if the goal is to have agents partnering with human beings that are your employees, there has to be a synergy with HR. So it's not just a technology job anymore. It's a business job but it's also an HR job, because to your point about culture. I found that really interesting observation that was made there, and it was all based on what he sees with clients, so we'll see where this all heads for sure, and we'll be tracking it. All right, let's move on before we wrap up here onto a technology roadmap, some thoughts from you guys. And John, why don't we start with you, just on your 2 cents on the big technology focus areas, and then Dave, we'll go and you can summarize what you've been doing with George Gilbert.

>> I think the technology area roadmap is going to begin that shim layer between the infrastructure, microcode, assembler code, real route coding, to that agentic layer that's going to enable all the vertical agents and all kinds of agents stuck in an agent. So I think there's going to be some middleware that will need to be written that has a systems construct to it that matches the acceleration of what's going on in the infrastructure. So to me, I look at the roadmap very clearly as what is enabled in the middleware layer, of the software layer, the data layer will be key. And I think everyone knows that data feeds AI. They just have to get their act together to what Dave was saying earlier and what the research is showing, is how you organize that is completely different than it was before agents came in. So I think the technology roadmap will be this data layer, harmonization layer, semantic layer, control plane, whatever that turns into, that the role of data has to be horizontally available and vertically specialized. We've been saying this on theCUBE for almost 10 years now, Dave. You remember. This is it playing out in realtime. I think that's the key. And again, speed to get the apps out, to test, because that's going to be key, but then scaling it will be super important. So if you're not thinking about scale and systems, then it's not going to work. That's coming out of most of the data we're seeing.

Scott Hebner

>> Dave?

Dave Vellante

>> So you're going to appreciate this, both of you guys. If you look back on the last decade, we've been preparing for this moment for 10 years. If you think about the four horsemen of the last decade, cloud, mobile, social and big data, cloud, mobile and social were transformative. I would say they exceeded their expectations. I would say big data was the one that didn't, and it was kind of a failure if you look back. I mean, yeah, companies like Snowflake and Databricks came out of it and there certainly were some companies who were able to apply big data, and Hadoop and so forth, and gain competitive advantage, but broadly speaking, most companies don't have their data act together. So when you think about the road to agentic and what that looks like, and specifically the building blocks to get to agentic automation, it's got to start with getting your data house in order. And Andrew, if you bring up this funny slide, this Wizard of Oz slide with the road, the yellow brick road, there's a lot in place, particularly with cloud. People say, "Wow, this is moving so fast." Well yeah, it is because you've got the internet built out, you've got cloud, you've got mobile. And it started in the bottom here with the cloud data platform, separating compute from storage. That is a good starting point, but the point of control is moving up the stack from cloud databases to these governing catalogs, and it's especially important as these things are opening up, like Unity and Polaris as examples. So the technical and the operational metadata become really important to be able to govern that. And the other thing is the data mesh movement has been important but slow to implement, but the whole idea being domain ownership, data literacy, data as product. Getting companies thinking about the people who have the expertise in the organization actually owning the data and democratizing that data and expanding that data literacy, and then bringing in business logic and process knowledge. So today for instance, okay, you hard code into microservices, if this then that, but they're going to be exceptions, but they're hard coded today. How do you make business process malleable and building blocks for the future? That is going to set the pace for true agentic, making those processes to be basically built on the fly based upon the situation, and the agent's able to do situational analysis and learn from exceptions and the reasoning traces of humans and get smarter over time. And so, thank you Andrew, the point is this is going to take some time to evolve. Personally, I think the better part of the next seven to 10 years, and it's going to be a journey. It's going to start with digital assistants that accomplish tasks and can take human instructions. RAG-based chatbots is a good example. Then you move into reasoning and thinking and multi-agent systems that collaborate with humans, and agents that are governed but also can learn. And then ultimately, to me the holy grail is you could take top-down goals of the organization. We want to grow revenue by this percentage. We want to gain market share, but we want to maintain margins. Put in those constraints and have agents that are smart enough and collaborative enough to actually create those processes on the fly based on the market conditions to make that a reality, and that is not going to happen overnight.

>> Yeah, I think one of the things I'll throw out there as a haymaker, I predict that here on theCUBE, that we're going to see an app store for agent models soon. It's going to be in e-commerce. We haven't even got into the interface side, which is going to be voice and/or composability. We're living in a Lego block engineering era where you got code assistants and essentially business logic at the top of the stack. So that was normally, Dave, in the middle of the stack, you remember? Go back to the old web services days, business logic was in the middleware layer. Like you said, it's moving up to the top of the stack. So to make that realtime, your chart is beautiful, I love the yellow brick road. It will play off the Wizard of Oz, which we saw at Google next. The man behind the curtain is an agent, so the wizard is an agent. So we have to see model integration. You're going to see agents looking for an advantage. Who doesn't want to be smarter? I think generative AI and machine learning pivot to the top of the stack where you're going to start to see realtime analysis in the agent layer. So I think the data layer is going to be a wide swath of intelligence, and I think that to me will be key.

Dave Vellante

>> And John's point is really an important one here, is the business logic historically was hard-coded. Humans, hard-coded, that business logic, and then the software ran it. And we're flipping that model around where humans are speaking to these agents in natural language. The agents are smart enough to understand the conditions and create the process essentially on the fly that maps to the situation as it is in the market, the organization, the skill sets, etc. And that is magic, and that doesn't happen overnight.

Scott Hebner

>> Yeah, we're definitely on a journey here and I think the yellow Brick road shows that too, agentic processes on how you run a business, the workflows. And as that journey unwinds, we'll be doing more and more summits that track to that roadmap there. This summit was at the very beginning, as Dave said, which is this is a progressive journey. You start and then you build from there. And let me just bring up the chart on building an AI agent. So we're talking about individual agents, because ultimately, you're going to wire these agents together, and this is what we talked about, right? Step one, you've got to start on your generative AI and LLM-based infrastructure and then deploy the domain knowledge models, and that's where you start getting into the semantic elements of your data across an organization and how entities relate to each other and the context that you have among them, very domain-specific. And then from there, you need to build models that can infuse explainable, understandable decisions, and that will allow people to actually interrogate the decisions and say, "What happens if I do this?" And do counterfactual analysis and things of that nature, where they can interact with it to really test out the decisions before they take action. That's a model ecosystem expansion on top of generative AI. Then from there, you build the agents. You acquire the ones that you don't want to build, you customize them, you integrate them, and then you start to build these and deploy these real-time agentic systems that both Dave and John were talking about. And then the whole system learns from each other over time, and you guys are right, you were flopping the equation, right? Build what you want and the code gets generated versus the other way around, and it's going to be a fascinating journey. So, all right, Dave, John, I really appreciate you guys being here. Some really great insights, gives us a lot to think about. For our audience, make sure you watch all 12 sessions that make up the AI Agent Builder Summit. Each of them touches on a different aspect of the agentic AI journey. From them, you'll gain valuable insights and practical advice from those that are implementing real-world solutions today. You can access all these sessions by visiting the summit portal on theCUBE.net, or by checking out our YouTube channel. And by the way, on that portal, you can also learn more about the participating companies and the leaders speaking across our sessions. And finally, stay tuned for our next summit in our series on agentic AI coming in late 2Q. Thank you so much for tuning in. We are the leader in enterprise tech news and analysis. Bye for now.