2026 ZKast - Taming the AI Flood: How Fabrix.ai Solves Enterprise Scale & Modern CVE Threats

[00:00:05]Well, this is GK here from GK from GK Research and I'm inside the world of solutions at Cisco Live 2026. I'm at the Fabric AI stand here in the solutions village.

[00:00:15]>> AI village, yeah.

[00:00:16]>> Yeah, I guess that's what they call it.

[00:00:17]I'm at with Shailesh Manjrekar, CMO of Fabric AI and Rashid Blouli, distinguished engineer. And so, how's it going, guys?

[00:00:26]>> Good, good, good. Well, thanks. Thanks for bringing us here, okay? And I'm glad to be here.

[00:00:31]>> How's the show been for you?

[00:00:32]>> It's been good. It's been good, yeah.

[00:00:33]So, there's a lot of AI conversations with customers, analysts, right? And partners, particularly.

[00:00:38]>> Well, it's all AI.

[00:00:39]>> Yes.

[00:00:39]>> Right? And so, obviously you've been a partner for Cisco for some time now.

[00:00:44]You've had an agentic solution that works with them. So, what are you showing at the booth this year?

[00:00:48]>> Yeah, so primarily we are showing two things, right? So, one is you know, it's showing autonomous ambient agents for primarily all the operational use cases, whether it be IT ops, sec ops, or collab ops, which is very interesting and we'll be talking more about that. But it's >> You don't hear collab ops very often.

[00:01:05]>> Yes, yeah. I mean it's I mean basically choose your unified communication interface of choice, right? Webex, Zoom, whatever, and you can invite a Fabric AI agent and provide voice commands, right?

[00:01:16]And it will do a lot of root cause analysis and things like that in the background. But that's one part of the problem or what we're showcasing. The other is a enterprise platform itself.

[00:01:25]And the platform comes with all the guardrails, right? Particularly film ops capabilities because you don't want your agents to kind of be rogue agents and consume all the tokens and the budget, right? But it also comes with all the observability, explainability, right?

[00:01:41]And so on and so forth building. So, our platform primarily it's it's purpose-built, right? It's meant for real-time data consumption. It is actually full stack. That means we go all the way from data integration, data enrichment, putting it into a central platform, and then running AI agents on top of that, right? And the third thing is it comes with all these guardrails.

[00:02:02]And there are three key pillars of the platform, agentic data federation, right?

[00:02:07]Content content engine for cross-domain orchestration. And the third is unified communications. And I'll let Rashid talk more detail about the three three key pillars of the three key modes of the platform.

[00:02:20]>> Yeah, so picking up on that, Rashid, there there's there's a lot of agentic AI platforms out there.

[00:02:24]>> Yeah.

[00:02:25]>> And so talk somebody says, "How are you different than those?" What's your answer?

[00:02:30]>> Uh my answer would be that um it's it's to try to get the agents to actually operate in a large-scale enterprise environment. Uh where the the demands on the agents are real enterprise demands. And that's what our our platform is is meant to do.

[00:02:44]>> Yeah, enterprise scale is much different, right?

[00:02:46]>> Yeah. Yeah, so things tend to fall down on the POC. And we spent a lot of time trying to figure out how do you make this work at scale uh in a real environment. So the three pillars that Shailesh mentioned, uh one of them is the context engine. That's how do we condition the signal that the LLM receives so that it's able to operate efficiently in large uh long workflows operating on large amounts of data.

[00:03:09]>> So the data is the what, the context engine is the why. Is that everybody think of it?

[00:03:13]>> Yeah, the context engine is going to uh put the right uh information in front of the model so that when you're completing uh doing completions in the model, it's able to complete the right pattern.

[00:03:25]>> Yeah.

[00:03:25]>> So we set the pattern for it to complete.

[00:03:27]>> Okay.

[00:03:28]>> The other thing is uh something that's called agentic data federation that we use to build what's called an ontology.

[00:03:34]So an enterprise ontology. So rather than having agents having to explore the entire data landscape inside of a enterprise environment, this is kind of a live map of where the data is, what the systems involved are are going to be, and what their capabilities are.

[00:03:49]>> I think small.

[00:03:50]>> And then lastly, this hyperconnectivity aspect of it is the ability for the agentic platform to connect to any system that exists anywhere inside the environment, regardless of whether it has MCP or whether it's a classical more legacy API type of system.

[00:04:05]>> All right. Now, Shailesh, I think it's about a year and a half ago that you launched Fabric AI and went GA with the product, right? Or at least rebranded the Fabric AI. What's been the How's customer traction been and then what's been the feedback from the customers in that time?

[00:04:19]>> Yeah, no. I mean, so we've been ahead of the curve, right? So, and we have real paying customers. Like, for example, we recently won an enterprise Fortune 500 customer, right? It's actually jointly with Cisco, right? But, primarily we're able to kind of kind of wow them with the agentic capability. So, almost 80% of the value of the agents they were able to derive in the first 6 months, right? And then we actually handhold the customer journey all the way. So, we have this concept of the FDE, which is now becoming very popular, right?

[00:04:50]Forward deployment engineers, who are who act like your trusted advisor throughout the journey, right? So, that's on the enterprise side, but we're also getting a lot of traction on the Telco and the SP side, which has been our forte, right? So, we we tend to work with large Telcos and primarily across different domains. So, that has been our sign of secret sauce, all right? Where we are able to integrate disparate data sources with enrichment, whether it is campus, data center, optical, mobility, and what have you. And then run AI agents on top of that. So, that's the enterprise customer traction. Similarly, we're also getting traction on the analyst side.

[00:05:27]So, we were recently featured in three publications, which Gartner put out just in Q1 2026, right? Yeah, and then lastly, right? Recently, AWS picked us up as the only agent tops platform for their MSP program. So, there are about 275 MSP partners for that program, and we are among the 10 tools, but we are the only tool which is actually agent ops tool. So that kind of talks to our kind of maturity and our credibility what we have been able to build in last 1 and 1/2 year.

[00:05:56]>> Yeah, no.

[00:05:57]Congratulations on those. It seems like great traction.

[00:06:01]At the show Cisco actually launched its own agenda ops platform. It's been working on that for a while. How what's the relationship with you and that product? You work with it, competitive with it? How's it different?

[00:06:11]>> Yeah, so that's a good question, right?

[00:06:13]And we've been a long-term Cisco partner, right? I mean and we have multiple swim lanes across Cisco. Our visions are very aligned, right? And you know, we complement actually the C3 announcement from today, which is a Cisco cloud controller, right? So the building blocks for C3 are essentially Splunk is the machine data lake or data fabric and thirdly data federation, right? And we already have an integration with Splunk where we complement them by bringing in all multi multi-domain disparate data sources and enriching that before we put that data into into Splunk. But at the end of the day, right? We main capabilities what we have is multi-vendor multi-domain, right? So we give the choice to the customer. If it's a Cisco shop, they would definitely be interested in going the C3 route and we'll complement that play, right? Or that sales motion. Whereas if it is a non-Cisco player, right? They may want to choose our agents, right? Which they see valuable. So that's the kind of play, but we we tend to work obviously Cisco and Splunk very closely, but we also work with other OEMs as well.

[00:07:17]>> Yeah, so you're not you're not trying to replace a Cisco cloud controller. You work with it. In fact, in a lot of ways you'll have customers get more value out of it.

[00:07:25]>> Absolutely. Absolutely, right? So yeah, I mean even the agents can work on top of the C3 federation, right? Or the marketplace what they they announced this morning.

[00:07:34]>> Yeah, no. It sounds like from what Rashid said, you you certainly sell to large enterprises, right? If if scale and the proof concept is important. Who are your other customers then what are the problems?

[00:07:43]>> Yeah, so our primary customers are obviously enterprise customers, right?

[00:07:48]So our guys like you know, our large enterprise and across verticals. So we are in healthcare, we are in manufacturing, we're in fintech, but also large SIs, right? So we are working with large SIs like you know, several of them, right? So Infosys, Capgemini and so on. And they are now feeling the pressure with all this FDE and the likes of OpenAI and Anthropic getting into that that business as well. So that's the second bucket. And the third bucket is obviously with OEMs, right? So we tend to work with obviously Cisco closely, but also IBM Fusion and so on and so forth, right? So that's kind of our three customers.

[00:08:24]>> Yeah, so it's really the big companies networks that have to scale and are complex and you help make them so.

[00:08:30]>> Yeah, and then also we are also available in AWS marketplace if somebody wants to get it from the cloud perspective.

[00:08:36]>> Uh plus they can burn out their credits that way, so.

[00:08:38]>> Yes, exactly.

[00:08:39]>> There are one one of the subtopics here I guess the show is Mythos, right? There's also OpenAI Daybreak and I think it's fair to say that um you know, every time something like that's announced it creates a lot of chaos in the industry. And so how is your vulnerability exposure agent helping to address the vulnerabilities created by these things? So it's I think there's a lot of uncertainty right now.

[00:09:04]Everyone's interested, but they're not really sure how to move forward.

[00:09:06]>> Yeah, so everybody's happy that we're discovering security vulnerabilities.

[00:09:10]Nobody's happy about how fast they're coming. So it's a bit too much for a human being to handle. It turns out that an agent is really good at being able not only to handle the incoming vulnerability reports, but then to also figure out are we vulnerable?

[00:09:24]Are we exposed to this vulnerability? So our agent is able to actually traverse the infrastructure, communicate directly with that infrastructure and identify places where that CVE is applicable and where it could be remediated if necessary, and then apply those reme- remediations automatically if possible.

[00:09:40]>> Yeah, I guess this is like uh the reason it was so hard to track as humans is because there's just too much data. It's too much. But your agent can literally see everything.

[00:09:49]>> Yes, this was created by Miso's uh like generating so many new vulnerabilities.

[00:09:55]It's already hard to keep up with CVEs on a normal day. Today, it's a flood.

[00:10:00]Agents never get tired and they never get bored.

[00:10:02]>> Yeah, and it actually underscores the importance of AI from an automation perspective as well. I was There was just actually talking to a uh customer about this uh before I came in here, where um the the number of CVs Cisco's going to be releasing, you know, moving forward is going to be way more and and way faster than what they have in the past.

[00:10:23]And so, trying to apply all those all all of those in real time or very small windows doesn't even have practical from a human perspective, right?

[00:10:31]>> Yeah, and just to add to what Rashid said, right? So, Miso's and the likes of Daybreak from OpenAI, right? So, they are really frontier models which are going to be very transformative, right?

[00:10:41]And what we see is, you know, a reasoning model kind of collapsing all the boundaries across disciplines, right? So, earlier you used to have SecOps different, right? Where you're just looking at the vulnerabilities. You to have IT Ops to find out, "Hey, what's your like exposure to the that vulnerability, right?" And you had SOC Ops where you are looking at threat intelligence, right? Now, the reasoning model has to go across all these three dimensions disciplines, right? So, that is where we see our platform and our agents essentially kind of unifying the data. They're unifying the domains and moreover they're unifying the disciplines, right? So, that's how we see the traversing happening across, you know, how things are changing with these frontier models.

[00:11:20]>> Now, I I know Cisco's one of the big Cisco's go-to-market motions has been secure or secure act factory, and you really see a recently announced a joint architecture with them. Can you dive deep into that a little bit?

[00:11:30]>> Yeah, absolutely. So, we are actually one of those certified ISV on the secure Cisco security AI factory and that's primarily for a class of customers who are looking for what we call sovereign AI, right? So, it's a new swim lane. So, these are the set of customers who don't want any of their data or any of the any of the artifacts to actually get to the cloud. So, they want everything on prem. They want models to be run. So, we are actually certified on top of OpenShift and we work with all the Nvidia names and or we have LLM depending on what you have, but we're also an Nvidia Inception partner.

[00:12:04]So, we're working with their ISV team and in fact this booth was actually by invite only. So, there are six partners around the room here if you can see and we are one among them who are certified on top of secure AI factory and that's what has landed us here in the AI village.

[00:12:17]>> Yeah, and I want to go back to the the conversation we were having around that.

[00:12:21]When you think about the velocity with AI's working with, I think the one thing we know is we don't know what's coming.

[00:12:30]Right? And so, I've talked to a lot of customers about them wanting to future-proof their environments and so, if a new model comes in fact Chuck joked on stage that since he started the keynote three new models got released, right? So, something like that happens, can you help your customers future-proof their environments and so, whatever happens to come tomorrow, a week from now, a year from now, they can be ready and take advantage of it?

[00:12:51]>> Yeah, I mean our platform is model agnostic and we test thoroughly with new models as they come out, but the meat of the infrastructure, the things that make agents work at scale, the model is just one piece of it.

[00:13:02]There's an entire machinery and so, yeah, we're able to keep up with changes in the models and we're able to scale with the problem.

[00:13:09]>> Yeah, well, that's good cuz like I said, we don't know what's coming. And Jilesh, what are your your routes to market? How can customers you know, acquire your product?

[00:13:19]>> Yeah, so customers can acquire just the software and the AI agents, right? So, we have our direct end-user pricing, but they can also acquire us either on as an sovereign AI solution through Cisco or a partner. So, we work with a number of different partners, whether it be Presidio or WWT, right? SHI, Carahsoft, a number of those SI partners, right?

[00:13:40]And lastly, we're also available in AWS Marketplace or through an MSP who is part of the AWS MSP program.

[00:13:47]>> Okay, and I think you had a little demo you wanted to show, right?

[00:13:49]>> Uh yeah, so actually what I wanted to show was uh you know, so this is a launchpad what we have, right? And uh this launchpad is is where you can see things like you know, network health dashboard, right? So, these are ambient agents who are what we mean by ambient agents is they're proactive and they're running all the time, right? And they're collecting data. They're So, you see the single pane of glass, right? Across all of this disparate data sources.

[00:14:15]Obviously, we have a very broad coverage across Cisco, but it could very well be Arista, it could be Juniper, it could be HP, name it, right? And that's that's the power we kind of bring to the table, right? That this is really an a multi-age a multi-domain multi-data source platform which can cater to multiple different you know, different domains, right? So, that's going back to the launchpad, right? So, there's also other capabilities like you know, you can look at the AI spend, right? You can look at there's an evaluation module which at every step of the way it's evaluating the decisions which the this model is doing, right?

[00:14:53]>> Well, everyone's talking about token economics, right?

[00:14:55]>> Token economics, exactly, right? And we provide you all these capabilities, right? So, actually I'll show you, right? It's not just the the the environment and the agents, but we can also provide observability for any agentic application. Like for example, we have a >> [snorts] >> We have this agent catalog which can shows almost 40 50 plus agents, right?

[00:15:17]But then we also have an admin dashboard which shows you the observability here, right? You can look at, you know, the the models. So, these are the models we support. These are This is all the tooling. Uh by the way, one other thing which Rashid talked about was our UCP, right? The Unified Connectivity uh platform, which allows you to integrate with not just any MCP server, but any API, any device, which is direct to device, right? So, we have the ability to collect telemetry directly from an Nexus or a Catalyst and provide an runtime MCP wrapper, okay? So, just to show you this observability part, right?

[00:15:53]So, this is becoming very important uh because, you know, it's not just you know, so we can look at observability for any agentic application, whether it be cursor, whether it be cloud core, whether it is uh something like Lovable and so on, right? So, we have all the cost analysis done, right? So, you can see this uh by the application, by line of business, right? Uh by IT provider, and then also LLM cost analysis.

[00:16:17]Uh and by the way, we have all the alerts. So, if you are you exceed a certain number of tokens or a budget, you'll get alerts, right? And which can kind of alarm you or kind of stop those operations.

[00:16:28]>> first 6 minutes of using it, you'll be burned.

[00:16:30]>> Uh exactly, exactly, right? So, we [clears throat] have this across So, all the heuristic uh right? So, this is very important when it comes to an enterprise platform because, as Rashid said, LLMs is just is a means to an end, right?

[00:16:41]It's really the harness or the scaffolding around it, which determines the enterprise readiness.

[00:16:46]>> All right. Well, that was great.

[00:16:47]Anything else you'd like to add?

[00:16:48]>> Uh yeah. So, essentially, Collab Ops agent, what Rashid earlier alluded to, right? So, that's a very interesting AI agent what we have announced, right? And essentially, what that does is again, you can choose your Unified Communication Platform, right? Whether it is Zoom, MS Teams, whether it is WebEx, right? And we'll invite >> And most companies use at least three.

[00:17:10]>> Yeah, at least three, right? And and it's it's Think of it as as a as a as a war room, right? But here you are actually uh inviting Fabrics agents, right? And you can invite them and it's all with human control. So, it's not a rogue agent coming in. Only when you admit, you'll be able to join that meeting and through voice commands, you are actually asking the agent to perform some action. Whether it be think of it as a key, show me all my asset discovery or show me what's my currency on all these assets, right? Or show me the root cause analysis for this particular incident ID. Right, anything else you want to add, Rashid?

[00:17:44]>> No, that I think that covers CollabOps.

[00:17:46]There's another agent that did a digital employee experience agent, which is another cool agent we just came up with, which is basically almost like a self-serve for an employee of a large organization.

[00:17:58]You know, imagine that I'm I'm an employee. I'm having trouble connecting to my VPN or I'm trying to figure out if I've completed all my onboarding activities. We have an agent that is examining all the different facets of the of an individual employees experience of the digital landscape that they're working inside of and our agent is keeping track, proactively reaching out to users who may be experiencing problems and helping them figure out how to help themselves.

[00:18:27]>> All right.

[00:18:28]>> Well, thank you so much.

[00:18:30]Yeah, absolutely. I'm glad to be here.

[00:18:32]>> Yeah, so on behalf of Shelly Benigar and Rashid Blooy, I'm Zeus Kerravala from ZK Research from the Fabrics AI stands and thanks for watching. Give us a like and also hit that subscribe button and we'll see you next time on the next episode of ZK Research.

[00:18:45]>> [music] [music]