This discussion examines IBM's artificial intelligence operating model and strategies to scale AI across the enterprise, drawing on theCUBE Research briefing at One Madison. It explores real-time knowledge, agentic workloads, streamhouse architecture, hybrid cloud, automation and data governance to highlight operational and governance considerations for enterprise leaders.
Ritika Gunnar of IBM, chief performance marketing officer, presents strategic guidance on scaling AI. Gunnar discusses real-time data flows, agentic identities for observability and cost management, streamhouse architectures, hybrid cloud integration and automation, and they emphasize an outcome-driven approach and treating data as always-on. Gunnar identifies four essential components for successful AI at scale: real-time data, AI-infused workloads, automation and hybrid cloud. Gunnar notes IBM orchestrates these components alongside existing customer technologies. John Furrier of SiliconANGLE and Dave Vellante of SiliconANGLE guide the conversation and probe practical implications for chief information officers and engineering teams, and they underscore implementation and operational priorities.
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
IBM's AI Operating Model Series. If you don’t think you received an email check your
spam folder.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open the link to automatically sign into the site.
Register for IBM's AI Operating Model Series
Please fill out the information below. You will receive an email with a verification link confirming your registration. Click the link to automatically sign into the site.
You’re almost there!
We just sent you a verification email. Please click the verification button in the email. Once your email address is verified, you will have full access to all event content for IBM's AI Operating Model Series.
I want my badge and interests to be visible to all attendees.
Checking this box will display your presense on the attendees list, view your profile and allow other attendees to contact you via 1-1 chat. Read the Privacy Policy. At any time, you can choose to disable this preference.
Select your Interests!
add
Upload your photo
Uploading..
OR
Connect via Twitter
Connect via Linkedin
EDIT PASSWORD
Share
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
IBM's AI Operating Model Series. If you don’t think you received an email check your
spam folder.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open the link to automatically sign into the site.
Sign in to gain access to IBM's AI Operating Model Series
Please sign in with LinkedIn to continue to IBM's AI Operating Model Series. Signing in with LinkedIn ensures a professional environment.
Are you sure you want to remove access rights for this user?
Details
Manage Access
email address
Community Invitation
Ritika Gunnar, IBM
This discussion examines IBM's artificial intelligence operating model and strategies to scale AI across the enterprise, drawing on theCUBE Research briefing at One Madison. It explores real-time knowledge, agentic workloads, streamhouse architecture, hybrid cloud, automation and data governance to highlight operational and governance considerations for enterprise leaders.
Ritika Gunnar of IBM, chief performance marketing officer, presents strategic guidance on scaling AI. Gunnar discusses real-time data flows, agentic identities for observability and cost management, streamhouse architectures, hybrid cloud integration and automation, and they emphasize an outcome-driven approach and treating data as always-on. Gunnar identifies four essential components for successful AI at scale: real-time data, AI-infused workloads, automation and hybrid cloud. Gunnar notes IBM orchestrates these components alongside existing customer technologies. John Furrier of SiliconANGLE and Dave Vellante of SiliconANGLE guide the conversation and probe practical implications for chief information officers and engineering teams, and they underscore implementation and operational priorities.
As enterprises move artificial intelligence into production, managing increasingly complex workloads requires more than powerful models and infrastructure. Organizations need real-time data, automation and governance to support AI agents operating continuously across their environments. How can businesses bring these capabilities together while maintaining visibility, controlling costs and scaling AI across hybrid cloud environments?
Ritika Gunnar, chief performance marketing officer of IBM, joins theCUBE Research analysts John Furrier and Dave Vellant...Read more
exploreKeep Exploring
What are the cost, infrastructure, and operational implications for organizations as they increasingly integrate AI and always-on autonomous agents across their workloads?add
What are the essential ingredients (core components) of an AI operating model that organizations must adopt to scale AI?add
What is required to ensure AI agents can make better decisions as applications use an increasing number of data sources—including real-time and edge data—and how should organizations handle data processing and governance to support that?add
How do learning loops—where AI learns from human reasoning traces and exceptions—play out in practice, and what technologies in the stack support that maturity process?add
>> (INTRO)Hello, I'm John Furrier with theCUBE. We are on location for CUBE Conversation with my co-host Dave Vellante, IBM's coverage of what's going on in the AI at Scale world. We're at One Madison, the flagship office here in New York City. Ritika Gunnar here, Chief Performance Marketing Officer, CUBE alumni. Ritika, we just got briefed with all the top industry analysts upstairs here at your beautiful facilities. We've chatted in the past about data, but now it's AI at scale. CIOs are looking at transformation stories around their business models. This has put the AI data, AI value, and ultimately the economics at the center of the stage. So what's the big story on AI at scale? What are people thinking about? What's the top conversations?
Ritika Gunnar
>> Part of what we think is incredible is that as many organizations are infusing AI all across their workloads, all across their organizations to really drive value, the truth is as you are more successful with AI, so is the cumbersome cost that you associate with it and the associated unnatural state of what you can expect in terms of the consumption across your organizations. Just think about it for a moment. In the world of cloud, when you had cloud in the internet era, when you had workload spikes, you used excess compute that sat in your hyperscalers. Now in the world of AI, as you're more successful in using AI, especially as AI is taking action, you now have workloads that are continuously being utilized. And so you are incented to be always on because those agents are always taking action for you. And so what organizations need to grapple with is what does that mean for them in a way where they can actually truly be successful with the outcomes and have the underlying systems that enable them to be successful.
Dave Vellante
>> And a lot of those workloads were kind of lift and shift. You had experience within the cloud. Now it's totally different. You don't really have decades of understanding what the profile is and the signature of that workload.
Ritika Gunnar
>> Absolutely true, because now these autonomous agents are actually the ones who are taking those actions, that are the ones who are using your data. These now autonomous agents are actually a new type of workload that exists. And so with all of this newness comes a different set of requirements for these enterprises. You
John Furrier
>> know, the big conversation that we're seeing on the marketplace side and bringing it together with the new architecture you guys showed today is that there's a real adoption curve happening. The experimentations happen. Agents are in demand. Now you have the underlying infrastructure enabling intelligence models. You can swap out models, but now the workflows, the specialty, the domain expertise, that becomes the scarce resource. What are you guys seeing and how does your new architecture address that? Performance, the value, because this is not general intelligence. We check the box on that. We're now at how do I specialize it? What's the domain? A big part of today's presentation had to do with that. What are customers adopting?
Ritika Gunnar
>> You know, one of the things that we did was we went back and looked at a lot of our clients that are scaling AI in their organizations. We said, what are the ingredients that are essential to be able to scale AI in these organizations? And ultimately, what we saw ended up being the manifestations of the AI operating model. Now, the AI operating model is something that we within IBM believe is essential for any client, regardless of what kind of technology they're using. And of course, it also manifests in an IBM version. But let me go through that and tell you the core components, and then let's riff on that. So the first is, for you to have any kind of AI systems at scale, it has to have real-time knowledge or the most current or up-to-date knowledge, not just about what the LLM knows, but also about your enterprise and your organization. So data becomes the first element. Second thing is about workloads, and this is different than workflows and encompasses workflows, but think of applications now that are in this new world, the kinds of workloads that you run that include those workflows now are infused with AI. What are you gonna do in an ecosystem where now on average in the next few years you're gonna see the average enterprise have 150,000 agents to manage? Okay. What does that mean when it comes to multi-agent orchestration to actually drive an outcome? So workloads become essential. The third thing really is automation. So now we see a world in which there's a convergence happening of how do you have agentic identity, identity and access patterns of these agents, including how you actually have observability and optimization of this entire estate. And how do you do that with now instead of FinOps worrying about cloud costs, you're now worrying about what it means to manage the costs of these AI systems. So the third area is really automation. And the last fourth one is truly hybrid cloud. Your data, your workloads, they don't move to a central place. You need to be able to operate these agentic workloads where they exist. So hybrid cloud becomes important.
John Furrier
>> You know, Rob Thomas gave a talk and then you came on after. The one thing that stood out to me was you guys hit the market shift. So the operating model, totally consistent with what we're seeing with these AI factories, all these large-scale systems, because they're operating systems, they're systems. The second thing that came out was that the shift is about AI-native-like environment where software was built for people, but now software's being built for other software. I think you mentioned, made that comment. This is the game changer. So just riff on that piece that it's not yesterday's cloud data stored there. You had the Streamhouse, which is an always-on version of, I guess, a data lake or a lake house. But so you're starting to see these shifts. What were the key shifts that people should pay attention to?
Ritika Gunnar
>> So one of the things that's really critical when we think about data and how these new agentic systems use this data. Think for a moment on agents that are making extremely critical decisions, whether those be on how you're actually creating financial transactions or healthcare decisions that you're making in an instant. You need that data to, one, be as current as possible, and two, that data needs to come from a number of different sources and be able to come together in a way where you're really putting context together. So What becomes important is, one, is your data fresh? Number two, is it holistic and a complete representation of what you need to be able to make a decision? Data that you may have in a lake house may be, for example, current, but it may be insufficient in that it is not all-encompassing of everything that exists today. So when we think about data, it's all of that.
Dave Vellante
>> So could we come back to the four ingredients? You got real-time knowledge, which is— that's where the data is, the workloads, which includes workflows and presumably processes is in there. You got automation. So Agentic Identities as an example, that's where you're going to control your cost. And then hybrid is really making location— not making location a blocker. That's right. Let the data live where it is. Okay. So I'm envisioning a nonlinear production line, sort of just-in-time manufacturing. So this operating model is so intriguing because that is where companies they're obviously going to get alpha with their data and their processes, but also going to get alpha with their operating model if they can move faster than their competitors. What are you seeing with your clients in terms of them being able to leverage that operating model for competitive differentiation?
Ritika Gunnar
>> I think you have to encompass all 4 ingredients to truly be successful and scale with your AI. Think about it for a moment. They don't all have to be from one provider, but you have to have all 4. If you have your data and that data is feeding your systems, but you're not really taking care of automation in a way where you have complete observability of all of your workloads with that AI infused, then you're not necessarily having an optimal way of picking the right model or picking the right agents for those kinds of workloads. And so there's a worry there of you may not be able to get the outcomes that you need. That's just one example. It could be the same thing as you create these new kinds of workloads, You need to be able to integrate with your existing systems. People aren't going to run away from their SAP systems that exist, from their Salesforce systems that exist. What does it mean to integrate with those and have workloads that are representative of your past and your future together?
Dave Vellante
>> And so IBM is the orchestrator of these ingredients or the chef. How should
Ritika Gunnar
>> we think about where IBM plays? IBM has capability and technologies in each of these 4 areas, and together they form the IBM AI operating model. But we know that clients already have existing solutions either in their data spaces or in their workloads that they already have, or in their automation technologies where we can leverage what they have and be able to build around that from an IBM perspective.
Dave Vellante
>> So ultimately they're creating their own operating model that's unique to them, but they're utilizing IBM and potentially other technologies to do it.
Ritika Gunnar
>> And our belief is as we're putting that all together. So you heard today about the real-time stream house capabilities that Confluent has and what we're putting out into market that we believe that is best in class. It is the most complete that is out there. You can build your own using open source technologies if you want, but we know we have something that is extremely efficient for you going forward. You can build your own workload management capabilities, but we know that we are building something complete for an environment and then we're making all of these pieces work together.
John Furrier
>> I love the— there was a chart up there. It said warehouse, lakehouse, streamhouse, and it kind of clicked because we always used to— oh, data warehouses are great, but that's just storage, all the data. Lakehouses, we love lakehouses that would allow you to do things with the data, the centralized model. But the streamhouse brings home the operating model. I want to unpack this because when we follow the constraints, which is entrepreneurship, new startups, innovation, or money investment, they go to constraints. The constraint is not the intelligence as much as it is the data in the model at the right time because it's a system now, not just search query results. You're seeing a whole nother interaction between these elements. Explain the role of the data because there's a lot of data. It's definitely abundant, but the scarcity is getting it right. Into the system.
Ritika Gunnar
>> Yeah, just think about it for a moment. The average application itself probably has 20 to 30 data sources. In this world of AI, that is actually gonna increase even further. And then when you think about the number of data sources that is gonna increase and the fact that a number of these are in real time, think about edge devices that exist out there and how much edge work is actually happening. Think about even the context from agents themselves. As you put all of this data together, what becomes really, really critical is that you're making decisions not only on data that— again, the warehouse is important, the lake house is important, but those are just a few ingredients to all the other data in the systems that are coming together. And so what becomes really, really critical is that you have the ability to be able to take all of these data sources and you should be able to process against them, be able to have governance against them. Our belief is if you can take data and if you can shift left a lot of the governance, a lot of the who has the right access to the right data at the right time, it becomes important because you flow that data in real time with the context and provenance of who actually has ownership of that data. And therefore the agents make much better decisions.
John Furrier
>> That's why the Venn diagram was the illustration, because you can be at the center of it. Automating data, all that stuff. And then out in the edges, you could just have data.
Ritika Gunnar
>> That's right.
Dave Vellante
>> And there's also an element of that data, which is, I think, the part of the promise is you can help harmonize that data to make sure you're having a single version of the truth as the bromide. That's exactly right. Ritika, generally speaking, there's been a lot of AI successes despite a lot of people sort of complaining about the lack of successes, but a lot of successes around data munging, data entry, cleaning up my CRM. That seems like AI is really good at that. It's that tacit knowledge, that tribal knowledge that we talk about all the time, that's a little more elusive. So how do you see that? And that involves the processes of the organization. Sometimes it's fuzzy. How do you see that playing out and how can IBM help?
Ritika Gunnar
>> Well, tacit knowledge, when you think about explicit knowledge, it's the knowledge that is already in our systems everywhere? And the tacit knowledge is the things that you and I know that it's kind of like that gut feeling of the things that we have done for so long. So it's the intuition that we have and how we make those decisions. A lot of that, we start to be able to even encode some of those in natural language or in our data. We're able to use that. Think about context today and how context is actually being fed into these agents as well. Context, if you look at even just prompts and how you would make decisions and how you're actually even giving AI systems a view of what is correct and not correct. That gives you a way to be able to start taking some of that tacit knowledge and really guiding as well. But there is always necessary, a portion right now because of where we are with AI, that human in the loop, the tacit implicit knowledge is essential. And this is why when you think about workloads today, you can't think about— there is a lot of it that will be autonomous in nature. But in most cases, if you're talking about your essential systems, you do have humans in the loop to be able to bring some of that tacit knowledge in place where the AI has not yet matured to be able to even get there.
Dave Vellante
>> So let's talk about that maturity process. One of the themes that John and I see, we go around to these conferences, the learning loops, I know, is a very popular theme these days. And so it's kind of the AI learning from the reasoning traces of the humans, maybe on an exception. How is that playing out and what technologies in that stack allow you to do that?
Ritika Gunnar
>> Well, this is when you think about the workloads and when you think about the data, this is why the AI operating model is circular in nature and why each of them are spheres because they overlap, because both of those things become really essential and intertwined. Think about agents as they reason and as they spew out logs and data around what made them make a particular decision and what decisions they actually executed on. All of that is logged and traced. So are the prompts that you actually input into the agent for it to actually execute on something. All of this data then feeds back as context into the data that you have where you need to actually then start to reason around it to make decisions that then feed back into how the agents make decisions.
John Furrier
>> I love this always on because to me everything's jumping out with Streamhouse, everything else, the operating model, it's always on, it's never off. So we're kind of entering an era because you need those loops.
Ritika Gunnar
>> And those loops need to be real time and always present and on to be able to iterate through that feedback.
John Furrier
>> Yeah, and it's definitely— that's a shift. I want to hear more of that. You mentioned earlier actions. I want to come back to the operating model because it's clear to me that's a value creation model. You mentioned actions. Agents are taking actions, humans are taking actions. Your new title is Chief Performance Marketing Officer. So the word performance is in there. The question comes in, what's the ROI? What's the value extraction? I'm the customer. I buy the operating model, makes total sense. It aligns with the systems trends. How do I put into action and how do I extract the value? Well, how would you— how do you frame it?
Ritika Gunnar
>> A lot of it starts with the outcomes that you want to go generate. So when we look at the AI operating model, it is about not just the AI piece, it is about the outcome that you're generating and what are the things that you're taking into account. And when we look at the AI operating model, it is the technology landscape of that. But in addition, you have to think about how do you re-engineer or relook at those processes from an end-to-end standpoint? And how do these technologies of data workloads, what we do in automation and hybrid cloud, really shift into that to be able to get those outcomes? I
John Furrier
>> mean, it changes in a way how you think about performance, because if it's always on the old analytics dashboards, as someone who's been close to all the data business at IBM and now architecting the new mission, how should people think about performance if they zoom out? How would you look at that if you were advising a fellow team?
Ritika Gunnar
>> Performance has a couple of different implications. Again, I'm going to go back to it has a connotation of am I getting the right outcome? So one, tell me if what I'm getting actually has any hallucinations in it. Accuracy matters. So for number 1, performance, am I getting the right data? Number 2, am I getting it in the timeliness that I need it? So if you give me decisions on yesterday's data when I need to make a decision based on the data now, it doesn't matter. So the timeliness matters of that. And 3, the cost matters there too. So if you think about how much did it take me to get an answer in near real time compared to what it could have.
Dave Vellante
>> How was that eval done? How does that— take us through a customer example where they have an outcome, they run a sequence and they stare at it and say, yes, this was right, give feedback to the system. How does that all work?
Ritika Gunnar
>> Yeah, that works very much so in some of our workload tools. What you do is you actually have an agent actually execute something. So as it takes an action, you have all the traces that actually happen. And what you can do is actually have multiple different permutations of that agent making a decision and having another set of eval loops actually doing an evaluation of did that agent actually make the decisions? Would there have been a more performant way, a more accurate way, a more cost-effective way to be able to do that? So in a lot of the tooling that we have, you can do evals. And I know you have a session on control a little bit later. You can do those kind of evals. In a way where as these agents are being executed, you can actually start to create some of those permutations and see what the output is.
Dave Vellante
>> And you want it to be consistent. That's the other piece. Absolutely. You don't want a different answer each time.
John Furrier
>> Well, Ritika, thank you so much for coming on. We see the hybrid cloud, distributed computing, check. Agentic, boom. Momentum, check. Now you got the data operating models for AI. What is your advice for leaders? Because people are trying to figure this out. I'll— we'll end with you playing out here.
Ritika Gunnar
>> I'm gonna tell leaders to be able to start with an outcome-driven way and start in a way where when you look at the transformation that you need to make with your data, make sure you have access to all of your data, with your applications and your workloads, make sure that you have control over those, with the automation set of capabilities and observability, make sure that you have all of these elements and that you're really thinking about the outcomes and you're willing to start with what it means to take a look at the business process transformation, cuz that's really where it's at.
John Furrier
>> We're seeing the transformation competitive edge emerge. That's the operating model. Create the value, extract the value. Ritika, thank you for spending time with me.
Ritika Gunnar
>> Always a great time to spend with both of you.
John Furrier
>> I'm John Furrier with theCUBE with Dave Vellante. We'll be right back with more CUBE Conversations from One Madison, New York City.
John Furrier
Dave Vellante