From Reactive to Proactive: The Rise of Agentic Networks

So our presenters, we have a really fantastic group today. We have executive leadership from Kentik and from ServiceNow with Avi Freedman and Gab Menachem. And we also have technical experts, Danica Shei and Steve Meuse from ServiceNow and from Kentik. And, we have a lot to kinda discuss. So we're gonna kick it right off, and I'll hand it over to Avi Freedman for market insights and trends. Awesome. Thank you very much, David, and thank you, Gab and Danica and the entire ServiceNow team for making this possible. Thanks for everyone, to everyone for joining. As we look at the market today, talking to global enterprises, what we're hearing is that they're managing, well, global and hybrid infrastructure. The world's gotten more complex, but the business still expects the revenue to flow, everyone to be able to coordinate, communicate inside the enterprise twenty four seven. It's become harder and harder as the infrastructure gets more complex to hire experts to do this, many of whom are retiring and vanishing from the world. But the pressure is to make the infrastructure run a hundred percent of the time, but with fewer resources. So what are the approaches that you can take? The key to addressing these challenges is really to have complete observability, including the network, of course, the infrastructure and the applications, putting it all into context, making it available. Historically, what did that mean? It meant making it available for humans to do the work. The great news is that it's not just type, that it's now possible to actually have platforms and AI help coordinate and supercharge operations for enterprises globally. Our focus has been on doing that in the network layer, which we'll talk about a little bit, but we're really thrilled to partner with ServiceNow to bring this to enterprise, especially because for the last five years, we've really seen ServiceNow, not Cisco or other network vendors, as the lead in automating IT and network operations, the human side, not the Python and and other scripts, but actually how enterprises are operating that infrastructure. So it's great to be doing this partnership. I'll talk a little bit about what the Kentik approach is as an introduction, to, the integration. If you go to the next slide. Our focus is has always been on seeing everything. All your networks, whether they're the ones you own or the ones you use, the cloud and the Internet, bringing that all together, and bringing the application context and integrations to that. But, historically, that's been humans operating these tools. Our focus in turning observability into intelligence has been teaching the systems and and agentic AI that we've incorporated how to do what humans are doing and actually focus on delivering, well, actual outcomes by using AI. Historically, there's been a lot of focus on event correlation, reducing signal. Our belief is that we can actually help extend and put under control of humans the entire find, debug, fix workflow cycle. So our efforts have really been focused tactically on a couple levels. One, make sure the platform itself has everything that humans and AI need to operate. So whether it's understanding traffic or metrics or performance, putting it all into context. And then because we operate a global platform with hundreds and hundreds of people running the digital infrastructure from service providers and cloud companies and global enterprise, we're teaching our systems how to do the same kind of operations that humans are doing. And as a result, we can actually enable always on monitoring and use the systems to trigger in and look at what are the real issues, where issues should be routed, and then bring humans in to confirm, do further investigation, and move on. As an approach, as I mentioned, observability is at the core. Agentic AI, which, of course, we've been hearing about a lot this year, is now available to help identify issues, triage, and take next steps. And where we're going with this is to bring basically a fleet of AI advisers operating the observability and turning that into intelligence with human oversight and integrated with the systems that enterprise are using, like ServiceNow, APM, and other systems. So, again, we've said agentic a few times. It's a new term this year, but it's real and being used by customers. Really three ways. One, we've taught our AI systems how people use Kentik and enabled autonomous operations within our platform. Second, identifying and routing issues and bringing experts to those right systems. As I said, ServiceNow is really the leader that we see, in the field, and, this integration, which we're talking about today, is a key way that, we'd like our enterprise customers to be getting value. We'll say this at the end in the callouts, but if you're interested in this and you're a ServiceNow, Kentik customer, please reach out to either company because this is something that you can actually use. The third way, which we've seen is, DIY. So basically going from basic APIs to, natural language interfaces that expose some of the same capability. A to a MCP is a little bit, down the road in terms of security and discoverability, and we can talk about that in the q and a. But we have customers already also pulling systems like ServiceNow and Kintik into their own strategies, And it's been really exciting to see because the pace of innovation, is really high this year, and it's not just hype. With that intro, I will turn it over to Gab, and, and the ServiceNow side. Thank you, Avi. It's a great privilege to be here today in front of, Kentik and ServiceNow customers and future customers. Since the introduction of GenAI into our lives, just a couple of years ago, we've been hard at work to incorporate this game changing technology into every workflow. So let me start with some inspiration from ServiceNow's own journey. And what I brought you today is is, this graph that shows how as ServiceNow grew to double and triple its size in employees, we didn't add support headcount in the same proportion. In fact, during the last year, we while introducing GenAI into all the workflows, we saw an opportunity to repurpose a lot of these employees to continue to build our platform. Automation and ServiceNow workflows, absorb the extra load removing repetitive tasks and things that, frankly, people are don't like to do. And the result is that every IT employee became more efficient, proving that automation can scale business far better than simply hiring more people. And as you can see here on the graph, we we stayed pretty flat throughout the years and even at the last, in the last two years, we saw a reduction of the number of employees we need in order to support our own employees. When you, look at the incident life cycle, identify, triage, remediate, validate, ServiceNow becomes the operating system tying it all together. By connecting with observability vendors like Kentik, we bring in domain expertise and context about the environment. That accelerates the triage, and we know the problem. We know who to involve, what action to take faster than ever before. That's the premise of what GenAI could do for your business, if you incorporate everything that, the observability vendors are doing into your ServiceNow workflow. So in the next slide, we're actually showing the full gamut of, what the operator and the AI system is doing from when something breaks until the remediation is actually validated. GenAI, amplifies each stage of the process. It helps identify issues by spotting hidden patterns, speed triage with smart recommendations, drive remediation with automated actions, and validate outcomes by confirming fixes and documenting results. Every step is an opportunity to inject intelligence that compounds into better, faster outcomes. So as you can see here, we've incorporated the these, AI, parts to the workflow, not just by using ServiceNow technology, but actually using agent to agent, communication, working with observability vendors like ChemTec in order to bring it into your ServiceNow experience. Now this doesn't replace the need for, an observability tool. It actually amplifies the value of it. So as you look at, Kentik as your network provider, understanding your network, helping you understand what works and what doesn't work and how to fix better, ServiceNow is able to work with Kentik seamlessly behind the scenes to help the first responder actually understand what's broken, what is the impact, who should I contact, and how should I communicate to them so it's more effective. And ultimately, looking at where this gets delivered in the ServiceNow experience, all of this work comes together in ServiceNow's event management console, what we call AI ops these days. Employees see the issue, the context, the recommendations, and the actions without leaving the interface. No switching between tools, no wasted motion, just a streamlined workflow that cuts the noise and drives resolution. We're very excited about the opportunity to bring this capability to market with leading observability vendors like Kentik. Many of our customers have multiple observability vendors even within the same service and are looking for the way to incorporate GenAI into their process effectively. I want to invite my amazing colleague, Denika Shay to show you the a demo of the workflow of the future and how Kentik and ServiceNow together make the vision of agent to agent workflows a reality. Thanks, Gab. Hey, everyone. I'm a product manager at ServiceNow, and I'm gonna show you guys a quick demo of how we integrate with Kentik. One second. Okay. So this is a demo of how, the ServiceNow ITOM product works with Kentik and how we're gonna bring network expertise into the hands of the first responder or l one operator through a agentic AI. So we're beginning in the service operations workspace. This is, what we call the central command center for l one operators or first responders. And this is a central hub for where all of the alerts are flowing into ServiceNow, and we can help reduce noise and elevate issues that require real human attention. So I'm impersonating an operator for a fictional online gaming company, and I'm responsible for the game studio service. And so that's a very critical application. So I'm gonna start by filtering down to that Gervasi, game studio, and I'm also gonna filter down to just critical and major issues. And so one alert jumps out to me. There is a normal traffic increase for the front end service. And so clicking into this alert, I'm going to be able to see a little bit more detail. So click into this alert, I can open up the impact panel, and I can see what entities have been impacted. This is the game studio, which I'm responsible for, and the front end service. If I click into probable root causes, there's also four probable root cause theories detected, and these are detected based off of recent changes that ITOM has detected, related to the Impact Gervasi. So just a quick glance, I can see that there's been recent deployments on the subscription service, the front end service, player matching service, and there's also been a recent TLS update, configuration change on the user service. So as a responder, there's enough information for me to act, but there's not a lot of information for me to act with confidence. There's a couple impacted services, and there's some conflicting signal on what could be the root cause. So there's a risk of going down the wrong path or pulling in the wrong team, extending my MTTR. So I'm gonna show you guys what we just released. These are our new conversational AI agents, and they're gonna help me, take more decisive action. And this is what we call analysis, which is ServiceNow's, intelligent chatbot. So pulling in in, my first question sort of off the top of my head is, is there abnormal traffic happening? And so what's happening in the background here is that Now Assist is getting to work, and it's grabbing contextual information about this alert, and it's going to assemble a team of dynamic AI agents to best answer this question. And pretty soon, it's going to detect that this alert came from Kentik, and so it's going to activate the Kentik AI agent. And I have it in a little bit of a verbose mode so you can see what's happening under the hood. But in the meantime, I'm gonna click into the dependency view, and this is going to show you sort of the topology map, in context. So the front end service is where the network abnormal network traffic is happening, and you can see that the front end service is calling into four different separate services, and these are the changes that have been detected on that service. If I scroll over to the knowledge section, we are also able to pull in, previous knowledge articles or postmortems that had very similar symptoms to what we're seeing now. And so in this case, you know, this describes a DDoS attack on the front end game service, and it kinda gives you some information on the symptoms that were observed and the root cause. So this is all very helpful information as I wait for this AI to respond. And so now assist has detected that there's significant network anomalies. There's been multiple traffic spikes and drops concentrated from one external service. And this is they're indicating that this seems like a DDoS attack that's affecting the game studio infrastructure. And so, you know, a a follow-up question for me is, can you tell me more about the traffic spikes? You know, as I debate whether, this is something that I can solve myself, or do I need to escalate immediately, knowing a little bit more detail? Kinda gives me more credibility as I pass this on. So, Kentik has detected that there's multiple traffic spikes. It seems like there's three. The first one happened in this window, and it went up to one point two four gigabytes. And these were you know, and there was also internal traffic surges. And so this is, and they keep kinda saying that this is consistent with a DDoS attack, which gives me more confident that this is what's happening. And then just out of curiosity, I'm going to say, do you have other root cause theories? What's the most likely one? And so based on the available data and the reports, the most likely root cause of this network anomaly is a targeted DDoS attack. It's indicated by the traffic spikes that it detailed for me. It also calls out, there's other possible root cause theories such as misconfigured network devices or sudden legitimate searches. I think it's adding that for completeness. But from here, you know, I'm pretty certain that I'm going to pass this on to a network expert to actually start the mitigation. So this quick demo just sort of illustrates how we can bring network expertise directly into the ServiceNow interface without having to pull in a network expert or logging into the Kentik platform itself. With just a simple questions, we can reduce triage time by highlighting the most probable root cause, out of multiple possibilities. And it helps me act with more confidence and clarity. And I can avoid unnecessary escalations to potentially, you know, all four teams. And so now I'm gonna pass this to Steve, who's gonna represent the network expert that the first responder pulls in to respond to this DDoS attack. Okay. Thank you, Danica. So as a network expert, on this fictitional network, I've I've been notified by the l one first responder team that there's an issue, and now it's my job to go through and verify the data. So, I might be coming into my alerting panel here to look for any alerts that I could see that I have some abnormal traffic alerts. So I can, go into the alerting panel, see that there's something happening, and even click down deeper into the actual alert that's happening right now. Now the l one team indicated that previous, you know, responses indicate that it might be a DDoS attack. So what this allows me to do is very quickly look at the traffic and make a quick decision whether I think this is attack traffic or not. If if I think it's attack traffic and I need to do something about that, mitigate the actual DDoS, I have quickies and easy options to, manually mitigate that traffic through a number of methods, whether that be through traditional hardware scrubbing platforms like a ten or Radware or use, something like a BGP flow spec or BGP communities to, scrub the bad traffic. But if I'm not necessarily convinced that this is an attack, I can look further. Right? So the Kentik interface has a mountain of data that we can start to look at it. So we can look at the pattern of traffic and is we we definitely see that there's something abnormal here. We can start to look at, things like which in network interfaces are the traffic coming in across. So we can see we have a gigabit interface here where some traffic's coming in, but maybe that's not helpful. Maybe I could know that it's open VPN traffic. I could look at the source countries and I could see, okay, there's a lot of traffic coming in from the United States, which makes sense as I have multiple branch offices, so there'd be a lot of normal open VPN traffic there. But But for these particular spikes, I can see that, okay, I have spikes coming in from for this particular type of traffic coming in from China and Japan. What I don't expect any open VPN traffic to be coming in from those locations, But we do have, people that travel, so we never know. So during this analysis, we can look at a a number of different aspects of the actual traffic itself very quickly, get understanding of services, packet sizes, and and all the potential attributes that the traffic could potentially have. But but the reality is is that there's there's an event happening, services are affected, and rather than make this a research project, which could could potentially take me an hour to find the exact detail, I have the option of using Kentik's ad visor. So I can very quickly hit the ask, a, Kentik advisor and say, what are the probable causes for this alert? And I can, maximize that to get a little bit better response here. And Kentik, AI advisor will start to go investigate the alert and give me some potential, things to look at. Now, again, the the real value that being delivered here is that, because there's so much data available about the network, we're trying to get services back online as fast as possible. So we need to reduce that time to innocence. Right? So here we can see, Kentik's already, the AI advisers already starting to dig into it, looking at patterns during that specific alert time frame, perform cause analysis on the traffic spike. So as this is happening, we can go through, look at the individual spikes, you know, do, is there anything unique about this particular spice spike? We can see where the which devices the traffic's coming in through. We can see that it's, what whatever type of traffic is it coming in through the DMZ? Is it not coming in through the DMZ? And there are multiple spikes. So are there issues between the different spikes? And is there any way to correlate this traffic? But as we glance down through here, we can start to see the most probable causes. So in this case, Kentik AI advisers telling us the most likely cause of this is a a VPN brute force attack. So, it might make sense for us to mitigate this traffic rather rather than spend another hour trying to build a case here. VPN service discovery scanning. So this could be somebody out in the Internet trying to look for open, VPN concentrators looking for security holes, or legitimate VPN client issues. Right? So maybe, maybe actually in this particular case, we built in, we deployed new VPN concentrators in China and Japan, and there's a misconfiguration. So every so often there's a as they're trying to connect or establish connection back to, the this main location, we're seeing these spikes in open VPN traffic. Right? So that gives you these options right away to say, here are the three things to look at to get your services back on, on track as quickly as possible. So in the case where I think that it could potentially be, an attack, I have the option of going back in and, you know, I've I've made a determination that this is a DDoS. I can come back into the alert, go back and select my my mitigation method, and and end the the service disruption for the time being. And then, you know, of course, in postmortem, go back and do a lot more investigation. So it gets us to the answer so much faster than it would before. In my time as a network operator over the last thirty years, you know, you can easily consume fifteen minutes just trying to understand how which systems are even affected. What system do I have to log into to start looking at the traffic? What devices should I be logging into? So this sort of just removes that entire process. Say we have all the data. We have all the systems. We know where this is all going. Here's the three things to do right off the bat, and and and get you to that, valued MTTI, the mean time to innocence. So in summary, being able to take this mountain of data between ServiceNow and Kentik and get to answers on both sides super fast, is the real value add here. Back to you, David. Avi, was this, your slide here? Yeah. I'm actually pretty, happy to see some questions that came in in the webinar chat and q and a that we'll get to, that talk about some of the that that are asking about some of the questions and benefits, that people get. But, what we've seen is, is, is some is some of what Steve said, faster resolution, both in terms of the time that humans take and, of course, actually getting the infrastructure back up and running, faster detection, noise reduction, more efficiency, all all in the goal of better delivery of experiences to users. We're excited about doing that, you know, on the Kentik platform and in partnership with ServiceNow, really going across the enterprise, from, from the first responders on up. So it's been very exciting. And, again, thank you to the, ServiceNow team for partnership. Alright. As you mentioned, we have a number of questions. I don't know, Gab or Avi, if you wanna jump into them or if you want me to to select some. We have some really good questions here from the the audience. Let's go. Hit us with them. Alright. Here we go. Interesting question from Bruno Ruiz, the very topical. It says, what is your view on the recent MIT research citing the success of AI pilots at five percent? What would you say are the key success factors needed to overcome pilots not translating into positive ROI and actually changing the processes and ways of working? So, Avi, I think I can I can start? From my own experience, both inside ServiceNow and working with customers on including GenAI and, actually, any kind of game changing technology like this, what I've seen is the, I guess the trend leads a lot of people to try to incorporate the technology as a technology seeking application. The the real value comes by thinking about how the this technology can change your operating method. So including it into an existing process or even changing the process in order to accommodate, the the benefit of that technology is really the thinking pattern that gets customers to be successful. So I think, I I certainly see a lot of our customers that have started using GenAI in different places and the excitement around using it. The places where they've been tremendously successful is when they rethought their process and incorporated the tools in the right or in the existing workflow rather than when they, are kinda, giving a tool to the employees and saying, you know, use it as you will. And what I I I I really believe that that five percent is in an added inflection point and going to change dramatically, over the next few months when all of the companies I work with at least are taking a much more strategic approach to using GenAI. And especially within the ServiceNow product, we're making sure that this gets incorporated into the way to take action, the way to investigate just like you saw in a way that feels very natural versus, you know, needing to go into any other place. And maybe a last plug to this is what you're seeing today is sort of a a way to use GenAI in the interface. But over time and even within the existing workflows that you saw today, some of this is going to be completely automated over time. So even with our September release at ServiceNow, the part where, Denica was asking questions, in the in the analysis panel is actually being, automatically, kind of started in the interface. So when you come to an alert, you would immediately see the first response of Gen AI looking at that alert and helping you triage instead of, needing to start that interface. So we're we're progressing into incorporating that, capability in a much more seamless way, and I think that's the future that will get us to much higher than five percent. I think from a Kentik perspective, maybe our sample well, for sure, our sample size is a lot smaller than MIT's. But, I look at, seven customers that we've talked to who've been working on AI initiatives across, we'll just say operations and infrastructure because that's the place that we play. And I would say that all of them have had some positive benefit. And so it's not the goal has not been to completely automate everything, but rather to bring together generally tooling and questions and and help people, become, you know, assist operations without needing to have that expertise. That's where I mentioned, I think, a to a and and, you know, it's a little bit early in MCP, but, just using APIs to bring together natural language help combine enterprise knowledge bases with some of the automation, I'll say people automation stacks like like, you know, ServiceNow is the leader and Kentik, in observability. We've seen positive success. So maybe I have not read, I I admit, the MIT article. Maybe I think the promises might be too big, but if you set the sights on, achieving progress and doing, the tractable, There's a lot out there that that people can actually get success with today, I think. Awesome. Let's let's jump in. There's a lot of questions, so we'll we'll try try to jump to go through these. An interesting one from, the the audience that an alert jumped out in quotes at the operator. Can your AI actually identify what alerts require deeper investigation? I'll I'll start here. Yes. In Kentik, for customers that have opted in, and this is something that's gonna be released generally, in early preview, very, very soon, You can go in and say, what what alerts should I pay attention to? What are the issues? And it will go and, the Kintec, AI adviser will go look through all the alerts and do triage and present the high priority and recommendations about what to do. And that's not just which one should I pay attention to, but what might be things that that I can actually do as recommendations, to resolve. So that's definitely something, that we've had a lot of feedback on, especially as, you know, customers often have a lot of alerts. The other one is, asking for advice on how to decrease the noise, how to increase the relevance, you know, as AI understands how to use the platform can actually make recommendations for that. And, to add to that, on the ServiceNow side, AIOps as as a technology really started as, the ability to correlate, alerts and understand better at what point do you wanna take action, prioritize the time of the operators. So the first value point you can get from AIOps is really this noise reduction that was called out in the slide before on in in the ServiceNow case, over ninety six percent of alerts are actually suppressed by not needing a different investigation or a human intervention. The other part that you can do is as we build these GenAI, capabilities and become better at identifying an issue, we can also build a remediation within ServiceNow that will automatically, go and fix the issue. And this can be used either with ServiceNow capabilities themselves or any third party integration, just like just like Kentik that we are basically calling from within the ServiceNow interface either automatically or at the request of the operator. So you really have a a a list of tools that you can use to take action, and that action is not only while you understand the problem, it can also be to understand the problem. So what you saw that Danica built as a as a capability to ask questions, we're also including in the next release in September, a form of a prompt that basically collects all the information an operator would like to know. Things like, what is the impact of this alert? Who should I contact to solve this? What what kind of, infrastructure network or application is, actually having an issue right now? How many customers are affected? All of these things that make prioritization easy and help you understand if that issue requires a deeper investigation. Well, going to the the next question here. And, if you ask the question, thank you for being patient as as we go through these. It says, can you please elaborate on faster attack detection? I think that's on your side, Avi. Yep. No worries. It's it's been a core part of the Kintik functionality that lives at the intersection of, of IT and security in that attacks on the infrastructure become availability problems. And so it's been, a big part of and focus of the Kentik, platform over the last years and something that over two thirds of our customers use. So as long as we have a traffic, ideally, metrics, if not, and then performance to be able to look at the impact of something that might be a spike, but if it isn't really affecting things, a lot of our customers are not that interested in, you know, prioritizing as something they need to go take a look at, it really should be possible to detect attacks that are affecting, critical infrastructure well sub minute. I think attack detection and remediation, especially on the DDoS side, as Steve talked a little bit about, is one of the most advanced areas for automation that we see our customers doing. I think as you see agentic AI, becoming more and more powerful and embedded, you'll see what we call the big red button. So AI surfaces something and says, hey. Here is a recommendation, something you can do about it. Click on this link to complete that automation. That's something we've seen on attack detection, to, you know, be able to completely remediate whether it's all on prem or with network routing changes or even engaging a global cloud scrubber. So happy to talk more. You know, you can reach out to Kentik. I'm Avi at Kentik dot com about kinds of attacks, but, really, I think this is not new news, but something that's it's possible to do pretty quickly and with good signal, on most infrastructures today globally. Right. And a quick one here. Do you have any use case studies for, automotive manufacturing? So we we definitely have a lot of, customers in in that industry. If you wanna learn more about the specifics of of their stories, you can go to servicenow dot com slash customers, and you'll you'll find a lot of these stories. Specifically, I think the the one that comes to mind immediately is we've worked with, an automotive, manufacturing company, in Europe that has has looked at the manufacturing, area just like an IT shop. So, basically, taking all the OT signals and the IT, from from the computers in the line and trying to find a way to manage all of it like a big IT infrastructure. And the efficiency that they saw from it was beyond just being able to address things, through one interface. They were also able to build, a framework to understand and and knowledge based articles and really predict issues because many of the issues actually came forward as log messages that were an anomaly. So when you think about the way to, proactively find issues, many of the signals that you wanna look at are actually hidden, and they can only be find through looking at, the low level signals like logs, like network, streams, and and things that mostly people don't look at because it's too much information to look at. That's where AI ops play a big role. And within the automotive industry, we have a lot of these stories, as I said, on the website that you should look at. I'm also happy to to discuss this privately, if this, if if if more information is needed, and I'll get my email at the end of this. And thanks. And, I know another question just came in about the same question for electric electrical grids and utility companies. And I think in both the manufacturing and and the utility side, we also have fewer than ServiceNow, but some customers there, some of them are are quite happy to talk to folks in their industries, but not, quite as open to talking in public use use cases, as they're running critical infrastructure that they wanna protect. But, you know, back to the automotive side, we see core IT as well as manufacturing as well as telemetry from the vehicles, you know, all three of those things that that people are using, observability and, and, and automation around, you know, in a growing way. Next question I can answer. Can we please get a recording of the call? If you can, we're happy to send that out afterward, and I'm glad I could, you know, jump in here. Alright. We'll go to something a bit more technical. How do you measure AI's impact on Kentik usability and performance? So it's an interesting one, because we've looked at when we release features, how do people engage with those? And, actually, in some ways, we wanna see more engagement. But as systems get federated, APIs become an important way of doing that. And as we incorporate AI capabilities, we want that use of that to go up even as the total time may go down that people are engaging with. So in the platform, we have ways of people giving feedback. We are for customers that don't opt out, looking at the internal usage of the Kentik platform in terms of the new AI capabilities as well as, you know, what questions people are asking that system can do better for. And so, you know, we measure that internally. And and and as part of our core product management focus, are people asking the right questions, getting answers, and and, ironically, is it reducing some of the use of some of the core slice and dice capabilities while at the same time, you know, the AI is actually using those on their behalf? So it's a different kind of product instrumentation. That's probably a different product management seminar for someone to do of, product management measurement in in the age of AI. Right. Here's a, a big question here. Network management professionals feel very cautious about exposing their data to AI to get guidance and to allow third party analysis of this data. How do you provide and guarantee security of firewall config and other sensitive network data? You know, how do you quantify the cost benefit results to your users? Do you allow customers to deploy your AI tech both on prem and in the cloud? So a number of questions there. I'll, try to split that up maybe into three questions. Kentik has the ability to deploy an on prem version of Kentik. It's still SaaS, that is managed. And in some cases, some of our customers, especially in the more regulated side, choose to do that. Separately, every customer has to opt in to using the AI capabilities, and, every single component of that that Kentik runs internally and with any partners, has a promise that no training is happening, you know, on that data. And you can think of it as all the AI capabilities being subject to the same limitations that humans are in terms of AI role based access. I'm sorry. APIs, role based access control, and what can access what. So you really think of AI as operating on behalf of the humans. It all starts with who's supposed to have access to what data. On the Kentik side, that's all multi tenanted and restricted, and we're really just extending that regime. And, you know, some of the capabilities that people wanna use their own, in house AI, is a little bit behind what we're doing in terms of the public availability, but that's all something that we're investing in, you know, on a continued basis. And I I am curious, you know, Gab, how is ServiceNow, you know, handling, you know, some of the same questions in terms of, you know, I mean, if we think about system of record, ServiceNow is that for a lot of our customers with what is where and what is the IT estate? Yeah. So I I think for for the security question, we definitely support on premise, in terms of, of where your infrastructure and networks reside. And the way we handle that is through what we call the mid, server, which is, an an in the DMZ server that collects that information and sends the the needed parts into the ServiceNow cloud server. In in a case of, you know, a complete air gapped, environment, there's there are solutions to that as well, although a lot less popular. On the AI sending the data to AI, I think, what you mentioned is is really key. I don't see a future where, enterprises are not using AI practically everywhere. And I think that as the industry continues to evolve this technology, more and more, kind of models would be, exposed into customers and and the ability to use ServiceNow specifically, brings today every capability as a as a as a way to use four different, market leading, LLMs. And you can plug in your own LLM if you want because we're built as a platform. So you definitely can change that into your own, AI infrastructure and basically not send anything, which is data into the cloud if you want to. I'm not seeing that a lot of the customers are actually this, cautious, and probably because it's the early innings of how to use this technology. They're mostly trying to gauge what the benefit would be before, basically putting it everywhere. So maybe keep the the, places where you have the most, important data to to, not expose to the world as, as a secondary action. But I think once you learn about the benefits, there's that cost and and, security, kind of chasm that need you need to go through. There's definitely a way to do that and a good way to do that even with, with customers in regulated industries. So I think that, that problem is is really getting solved across the industry, not just within ServiceNow and Kentik. Alright. We got we got time for a couple more questions here. Alright. Here we go. This is a good good question. So comp comparing between agentic ops and common network automation, what level of time saving for the analyst is expected? And also can we say that implementing agentic ops may allow the business to reduce team headcounts? I think what we've seen is I'll I'll start with the second part of that. Empowering the experts to do the job they were hired for. I think that's a a key underlying theme about the use of AI is not as much about reducing headcount as helping the tier one to self serve route resolve, helping adjacent teams to understand, is there a network issue so that it can be routed correctly, and dealt with by the network team if it is, and have the experts that were hired to do architecture implementation, engineering, scaling focus on that rather than on the more root tasks. So I think both sides of the first part of the question, both sides of that are important, both the agentic ops to watch everything, winnow down what's worth human, attention, make recommendations. It's important, and then you need some automation to to to connect. I think one of the great things about, some of the AI capabilities is it really helps network people under be able to use APIs to drive some of the automation, that people want to do. Although there's still some scaling challenges as we see a lot of our customers. Every engineer still has their own Python scripts for any APIs. There's still some systemic work, to do there that I'm happy to talk with folks about, what we see, you know, across our enterprise customers. That's it. So my vantage point, I think, Gartner kinda puts AI, in in two buckets of, domain expert or domain specific AI, and then there's generic. I think in the in the case of domain specific, AI where, for example, within Kentik, the ability to look at the traffic and understand just like a network engineer what's going on, recommend, root cause analysis and things like that. It's it's a great help for the network engineer to, better their performance and and do things faster and more accurately. I see that as an opportunity to, do things in a in a more cost effective way and and get a better result. As you climb up the ladder like like you saw in in ServiceNow with looking at full departments, I do see that customers even at the, at the lowest levels of the organization, but certainly at the leadership level, are looking for ways to, use automations and GenAI as a way to, you know, flatten the curve of employee growth in certain areas so that the growth of the company continues to happen while you have sublinear growth in the amount of employees you need to actually serve that customer base. I think that's a great metric for leadership to be calling out to the market. We certainly see the biggest companies in the world doing that these days. So the opportunity is there to, save money on things that required a lot of people, and now JennyI can do the mundane work for them. We still require, and that has been, very consistent across our customer base, that business decisions will be made by humans. Obviously, no one wants a a a network AI that will take down the network during, Black Friday or something like that because there's a spike in the network. So business decisions are still left to people, but a lot of the mundane work and investigative work that takes a lot of time can be faster and more accurate using these technologies. And, Gab, I think something you mentioned to me that, for ServiceNow customers that your IT group actually has a focus on talking to customers about how they're getting some of these kind of benefits because that was a really impressive slide that you showed. Yeah. And, so for ServiceNow customers, I think they can talk to you guys more about that. Yeah. Certainly. Let let me see if I'm able to understand this this question correctly. I believe it says, I use ServiceNow for opportunity management and for IT ops. Can I run any process like traffic engineering with observability integrated with ServiceNow? So I'm not sure I get the the full question, but let me describe what I understand from from this. As you look at, past incident insight ServiceNow, we have a product called AI ops leap, which is learning enhanced, automation playbooks, which allows you to look at, the opportunities to automate basically past incidents. So you'd see how many incidents are in a cluster and we automatically cluster them for you and say, for example, all of these, incidents could have been solved by this automation opportunity. And then you could take that opportunity and build a a full fledged automation inside ServiceNow or with third, party vendor and actually continue to build those remediations that could be completely automatic. Part of it is looking at, traffic data and and, observability vendors as inputs into that process. So not only looking at, the incident itself, but also the underlying, infrastructure, network, application data that brought you there so that you have a better way to identify that this is the actual problem to solve and that that the automation is a great fit to to solve that problem. I hope that answers the question, if I understood it right. Yeah. And I think, I think that as well, yes, people will be able to reach in and ask some of those super nerdy questions that people wanna do specific traffic engineering and such, you know, via the integration as well. But we wanna make it accessible, you know, via this integration to people that probably are not trying to, you know, change the routing, you know, necessarily right from that interface. I think we're at our our last question here, which is with the integration into ServiceNow, are you seeing more network teams using AIOps more for their main place to monitor for issues, or do they use Kentik or both tools to monitor for issues around the network? I think that really just depends. You know, network team is a big, is a big term. There might be twenty different subparts of network teams across, you know, three or four different teams. And so I think you might see architects still use Markantic, but there's so many people that are in the life cycle of the network, that, you know, we see. As I said, ServiceNow is their automation platform that that that routes selects and and that a lot of those network teams already operate in, and we're thrilled to, you know, bring them more functionality in that. I think, the way I can elaborate is that, ServiceNow serving more than twelve thousand enterprises, today is, seeing a lot of these enterprise workflows where there are different teams that are responsible to identifying an issue and then promoting it to the domain expert that can actually take action. In that case, many of the, processes that we see start with, a help desk or a network, or even a just a, a I would call it a a an operation center that really looks at all the alerts that come from different sources through our AI ops, reduce the noise, by as much as ninety six percent, and then look at those alerts correlated into infrastructure network, application, all the different stack elements, and then correlate it with data coming from observability vendors. What Danica showed today, her innovation is really being able to do this at the agent level, but the data level also lives within Gervasi now with the product we call service observability...