Unlocking the power of AIOps with Riverbed's Charbel Khneisser

Speaker A: Over the last two or three years, IT leaders and their bosses have been making very substantial investments in the latest and shiniest AI objects in the room. And they've been doing this because they have correctly seen that AI, uh, has the potential to completely transform how increasingly complex IT infrastructures can be managed. However, in the same amount of time, those people have often found that their investments didn't really generate the return they wanted. Maybe their data infrastructure just wasn't really ready for it. Or maybe they've picked a multiplicity of solutions that have ended up being deployed in different areas in a siloed way, and it's ended up with a rather fragmented result. So how do you do AIOPS properly? Today's guest on the Node podcast knows a thing or two about this. He is Charbel Neisser, the Senior Vice President of Global Solutions Engineering at Riverbed. Riverbed should need no introduction. They are, um, a preeminent leader in, uh, IT observability. And Charbel has got a few ideas about how we should be thinking about AIOps to make sure it works. Shoalbel, welcome to the Node podcast. Today we're talking about AI in the context of, of IT operations. Now, um, AI in that context is probably one of the most obvious things you can think of in terms of understanding what AI can do in its various flavors and the kind of needs that IT folks have got, um, to see what's happening across their networks and infrastructure and be able to respond to it and improve things, etc. Um, but that doesn't mean it's simple or easy or that people aren't struggling with it. So that's what we're here today to talk about. And it's great that you're here because you are definitely an expert in this area. Um, but rather than me, uh, you know, ah, attest to your credentials. Ah, I'll give you the opportunity to do that as well and see if you can do it humbly. Um, so tell us about yourself and also of course, your employer, Riverbed.

Speaker B: Uh, first, thank you for having me here. I'm really pleased to be with you today and have such a wonderful conversation about AIOps, which is talk of the town, wherever I go, at all of my customers, uh, about myself. So I'm currently leading the solution engineering team globally at Riverbell. And before that I had, in my 22 years of experience in IT, I was always focused on really helping organization, I would say, work on overcoming the challenges of performance. So since I started my career in it, I was more vested into Visibility tools, that's how they used to be called in the past and at the same time it all evolved to what it's known today as observability towards AIOps. So that's where my experience comes into my whole uh, career experience now. I was lucky as well in the last 12 years to be part of Riverbed. Uh, when I joined Riverbed, it was when Riverbed was shifting from being the 1 optimization vendor into what it's told the performance management vendor. They were more focusing on really helping organizations starting looking into what are the challenges that comes from it, whether on the network or the application side and helping them move from reactive approach to a proactive approach. So I acquired a great experience as well from Riverbed, which is today one of the leaders into the IOPS platform. From a background academical background perspective, I do hold two master's degree, one on the networking and telecommunication and the second into security of networks and system. And I do lecture as well at first year's master degree in security.

Speaker A: Okay, so um, as intimidating as it is for me to then continue this conversation knowing how overwhelming um, the ah, the imbalances between our expertise, um, I'll press on nonetheless. So um, tell us about. It's easy to sort of understand how AI can be so useful in the context of um, IT ops. And I bet you spend a lot of time sitting around tables with um, C level types who are reaching for those AI dongles and saying hey, that's definitely what we need. But it's not quite that straightforward obviously. What are the pitfalls and the challenges that people are Ah, that CIOs and their teams are facing right now when it comes to not just grabbing the shiniest AI tools, but implementing them in a way that really actually demonstrates value.

Speaker B: Yeah. So look at all the meetings that I'm having with C level or execs at uh, our customers or new customers existing, whether new, they all kind of share seeming concerns. Three main objectives. Number one when it comes to what's the major challenge they are facing today is the number of tools they have deployed in their environment. Funny enough, and surprisingly actually you're going to see that around 20 to 30 different tools exist in organization simply to monitor and supervise and provide information on where everyone are coming from. This by itself cause a challenge. Number one challenge from operation perspective, how to manage, operate and maintain those tools. The second challenge is the cost itself in term of renewing licenses, uh, managing the support, managing the vendor, the contract of the vendors and all this stuff. And it makes it Even more complex when problems happens, where should I go and look? Which tool should I trust more? So that's where what we call all CIOs are asking for tools consolidation today. And since AI is kind of one of the major aspirations, not only from CIOs but from business executive, believe it or not, CEOs are enforcing the application of AI, uh into the business. So that's becoming a mandate. But now for CIOs, whenever they are asked to do so, they want to have AI that works at scale with a strong ROI to justify the investment. Right. So when it comes to consolidation, they want to apply those consolidations on, consolidate all these tools, but as well penetrate AI into it. Now having said that, it becoming like they want to make sure they are able to adopt the AI, make sure it works in their environment, provide the value they need. But that's where the second, third I would say concerns they have are my data models supported into the AI algorithm I want to deploy. And becomes a question how good or how ready I am to get this AI implemented and to be operational. Because when they were riding the hype in the last years, many organizations failed in properly adopting AI. And now they are more skeptical to go back, but still they want to go back, but more prepared.

Speaker A: Right, and so how um, we had a conversation previously where we were talking about how um, proper value added, adding AI implementation from an OPS perspective obviously requires um, that you don't put garbage in and um, therefore don't get your garbage out. Um, but also that you've thought about the full landscape of infrastructure that you're trying to implement, things across and all of the different data sources. But then finally, and perhaps most critically in terms of adoption, there's the simplicity of it because obviously if all you're doing is making things more complicated, um, it's going to be trickier. So um, you had these three words, smart, uh, open and simple. So talk us through those three words briefly in terms of how you and Riverbed tend to view this topic.

Speaker B: Yeah, look, I mean going back to what I was mentioning before, the challenges that CIOs want to overcome today, one of them is how they make AI work and how much they are ready for those model based on the data they have already in place. Now when we start talking about really bringing the silos together by consolidating the tools, this comes to the fact how can I make sure I'm able to tap myself into all the domains that exist today in the infrastructure, in the environment or the empty landscape Those domains can be the applications, the network, the users, the various type of users, whether they are located in the branch office, whether they are located home, and the type of device they use. Because today mobile penetration is extremely high, right? Even doing work from mobiles, tablets, not only desktops or laptops. So this type of, the type of devices being used change dramatically as well. So looking at all this, being able to capture all this telemetry efficiently is a challenge. And consolidating the way you collect this telemetry again is one of the main objective of Sea levels by tool consolidation. So that's where Riverbed did it in a way where we introduced simplicity in how we collect all this information using a unified agent approach that can tap, that can sit on the end user, uh, collect network telemetry and user telemetry and the same agent can even tap itself to provide unified communication, uh, information about teams, uh, zoom calls, etc. And how it can integrate as well with third party tools like Intel Thunderbolt to collect information about intel behaviors on um, end user devices. Now that makes simplicity, uh, now whenever we're going to augment this data, going back to what you mentioned, if you're going to say garbage in, it's garbage out. This is all around the intelligence that you apply on top of this data. And that's where they want to have AI that work. So that intelligence requires first applying the proper machine learning algorithms to make sure we understand behaviors, baseline those behaviors and benchmarking the good to the bad. Right? And once that's done, this is where you start applying another algorithm, kind of causal AI to know where the problems are coming from. And then you apply expert system AI to automate troubleshooting and communication. Now all that won't be complete unless you integrate this whole ecosystem or this whole platform into, into existing systems and protect the investments that they have done. That's where openness comes in. Being open in a sense where I can really pull data from third party tools or really or maybe inject data using third party tools. And that's where Smart Autel or OpenTelemetry or ForeverBet comes in to expand that openness. It's like a parachute, right? So parachutes only works well when they are completely open, otherwise your landing won't be safe. That's one of the challenges I would say organizations were facing in the past by applying AIOps to a single domain. They failed because it was only again a silo based AIOps or the data that were collected were not highly granular enough to get the Proper algorithms outcome.

Speaker A: They were taking the candlestick approach, I think that's what it's called. Right, Candlesticking.

Speaker B: Um, yeah, I would say they were trying to fast approach things or going the shorter route. Right. And I would say maturity, it was more a hype before if you look at it, it was more of a hype where people were more rushing to get things in but they were not mentally prepared, operationally prepared and even knowledge, from a knowledge standpoint, ready enough to make it work.

Speaker A: Right. Um, but we're moving along that maturity curve I guess. And so um, in the end it's about ROI and maybe you've got examples for. I know you've made some announcements recently, um, that you might want to talk about, but you might want to also, uh, if you've got a couple of examples you can give to us of um. I don't know how much you can actually talk about, but I'm going to ask anyway. Um, successful applications of everything you're talking about. Maybe you can give us some idea.

Speaker B: Yeah, I have plenty actually of examples but I'll focus maybe um, I'll give you some good examples from M Financial sector because I know that relates to everyone. We all deal with banks and um, transact through banks. Right. One of the top five global banks in the world had for example a sea level initiative where it aimed to save $100 million in five years. Right. And uh, that's kind of everyone is being asked to reduce costs, right? Yeah, that's, that's, that's common. And the way they wanted to do that is by really having AI automation in place. So it's like leveraging AI to provide a good ROI of $100 million in five years. And they wanted to really make sure they are able to eliminate their kind of some certain IT activities that can be directly automated through AI systems and avoid for example some incidents to be routed to support calls or to the contact center and being troubleshooted by humans. And this is not replacing humans because that's distinct differentiation. We wanted those human beings to be focusing on more important tasks, more productive, more innovative projects rather than simply troubleshooting what I would call the obvious. Yeah, uh, so, so, and, and the first thing whateverbed was able to help that leading bank is to re. In the first three months we were able to operationalize completely that center that they had the contact center from a support desk automated in a way where we're reducing today the meantime to repair by 9 to 20% and save them like $1.6 million. In three months. So like that was a huge kind of ROI for them that directly was able to materialize in three months of automation using riverbed platforms that we discussed briefly before. And we're projecting like in the coming year we're going to be able to save like $32 million for them, hoping we can fast track them towards their $100 million saving initiative in five years. That's for example one, one of the top five leading banks. If I want to go more and focus a little bit on one British bank, which is as well a leading British bank who has actually quite big install base of users, they have around 150,000 devices where they are more kindly monitoring daily the uh, kind of the problems that are happening on uh, the end users or which is impacting their productivity, their user experience, uh, etc. So we're able to help them actually to proactively start using machine learning that I briefly explained. By leveraging the machine learning we're able to proactively identify problems alert on them even before the users start feeling it. That's where we shift them from a reactive approach to proactive, but as well to become more preemptive in a sense where we're automating some actions to remediate those problems before they exist. Today we have in the last year working with them closely on this, we have executed 322,000 remediation activity. That's kind of, if you Compare it to two man hours or two Ah, resources, that's kind of 7,000 hours of troubleshooting, right?

Speaker A: I was trying to work that out in my head just now actually. Yeah. And so this is proactively getting ahead of problems perhaps even before they've really become a problem.

Speaker B: Uh, yeah, 100%. Look, I'm going to give you an example. Consider driving a car, right? When you're driving a car, sometimes you get these system engines or it's going to tell you like okay, you need to check the oil, you're not yet in the problem, right? You're going to the problem, but you're not yet there. So what we're doing actually with riverbed even before we get so we reach this stage where we know if, if you're going to drive like five miles more, you're going to have this service issue. So we're even notifying it before the users start feeling it. Now already when you get the system engine or the engine check on your car, you already know there's going to happen a problem. What we're doing, even with riverbed we're notifying it before the users start getting notified that he might experience problems. That's where becoming more preemptive into this whole life cycle. And that's where AI, when it's well applied into practice, with the right data in place and the right collection, it's going to do huge things for you. You can move mountains is the right

Speaker A: way if you do it the right way. And uh, one of the things from my limited experience when I've been looking at Riverbed's modern suite of capabilities is its ability to communicate as well through a sort of generative AI function where even someone who's so high up in the organization that they actually have no technical knowledge at all can start to actually gain insight and knowledge by simply asking questions. Right. I might be over egging that, I don't know. But is that a kind of. It's not just simplifying the actual operations, but it's actually simplifying the understanding of them for a wider group of people as well. Would that be true?

Speaker B: Yeah, look, what you mentioned is totally true and I would say there's as well still in the market some confusion between chatbots and gen AI. Right. So let me go a little bit step backward. So what are the type of AIs that exist today in the market? Right, because that's going to help people understand the differences. So the first thing I would say is we have a type of AI which we call it cordial AI that's going to help users identify causes of problems. Right. There is an AI which is more like what we call predictive AI that's going to predict that there are problems coming in as I, as examples I gave. And then there's what AI we call a gen AI. The gen AI is what you describe. The capability to really go ask questions in a human language. It can be whether we are chatting or whether we are talking like as we're doing now, or rather for example by sending an email to an AI system and you expect an answer. But the difference is the conception is. And that's where there's a confusion in the market. It's not a chatbot because it's not like returning an answer, which is the obvious answer you can go read from a report. It's actually you're triggering an expert system AI in the background. So Genai to be called really Genai, it should have what we call the normal language interface on the system, but at the same time in the background enriched and empowered by expert system AI that can go analyze all the information that it has to return an answer which is not obvious for you to go find. Now that's what Genai is. And what we're talking about is not something that's a roadmap. If I want to take for example from a riverbed perspective. Riverbed announced Genai in the last three weeks in a big launch we have done along with other different launches we did. But Genai is today something we ship to our customers and we have many customers in the world who have been in the beta stage testing it and getting extreme results from it. Imagine we're sitting together and having this conversation now and I receive a message that for example our CEO David Donatelli is having an issue. M what I would do, for example in an unusual situation, I would have to cut it off and go figure out what's happening or what's going wrong.

Speaker A: Yeah.

Speaker B: Now with Gen, I can simply chat via teams with our IQ system, which is our AI platform. Ask the question what's going wrong with the team session for example and I'm going to get a response back by analyzing the multiple domain using the expert system of Riverbat. Analyze all this information and even be able to remediate the problem while I'm sitting on the call with you without zero interruption. That's innovation. That's where AI is put into practice to serve humanity and serve the business.

Speaker A: Right.

Speaker B: But evolve solution as well of Genai, it's going to be in the future what we call agentic AI, which is a new approach of as well leveraging AI capabilities into different type of ways.

Speaker A: Yeah. So I think um, just to visualize what you're saying is essentially we've got a parachute that's been extremely well constructed because it's capturing all of the air. And the CEO, uh, who sat on top of the parachute, rather dependent on it because if that thing candlesticks, he's got a problem. Um, but you can just. Because, because he's on the parachute. The parachute can communicate to him quite easily that there's a, there's a mountain coming up or whatever it is that he's about to drive into. Yeah, got it. So um, I think I've borrowed quite a lot of your time already. But um, let's, let's just say if there's, there's so much you could get into here. There are so many different, uh, well, so many different riverbed, uh, products if you like, or service areas that we could talk about. Um, we'll try and put some more information about that on the website along with this podcast. But if anybody out there wanted to know more, if they wanted to collar you somewhere, um, where might they be able to do that and find out more down the line?

Speaker B: So definitely they can go to, uh, Riverbed website. We have plenty of information that explains what we have spoken about, right? They can find out more about our collection techniques, right? How we can collect really the multiple type of data from all the domains, how we analyze that data using our AI, what are the type of AI we have, what is our view on Genai, how we do it, what's agentic AI and whatever it is doing there, because that's something we mentioned that we're investing a lot heavily on. At the same time, understand our data store capabilities. Because I would like to say as well, one key important point, referring to the point that you mentioned, it's all about how we collect, but it's not about how we collect, how we store. And that's one of the challenges that today's organization have. Uh, how can we store that data, how we can really put it at work? And our data store is extremely efficient. It's extremely powerful because it really use what I will call with tips that we have put like the observability at the heart of what we do. And that's why we synchronize everything that we have collected from all the layers in the digital ecosystem, right? So that's again, some important aspects that people can go read about on our website, right? How we align all these domains and we keep the data where it should be indexed as a source and we pull it on demand. At the same time, we do like our flagship event, Empowered X in October every year in multiple major cities. Like we're doing one, if I want to talk about Europe, we're doing one in Saudi Arabia, Paris, uh, London, and we do as well in many locations in Australia, Singapore and, uh, in us. So people watch out for that. They can go register, they can go attend, they will hear about all the innovation we're doing, we're launching, and at the same time they can reach out directly to Riverbet and send them inquiry on the website itself.

Speaker A: Brilliant. Well, it's an exciting kind of place to be, isn't it? And if you're going to be at a place for I think more than a decade in yourself if going to be lecturing people about this stuff, um, you're in the right kind of place, aren't you? Because you've got your parachute and it's really good. So, um, with that, I think I'll thank you for today. Uh, again, once again, come, uh, to the website. You'll find lots of links to all of the things you've just been talking about. And down the line, maybe it would be interesting, maybe after you've got through all of your events later in the year, it'd be quite interesting to catch up and ask you the same question again, which is, you know, what are people struggling with? And then maybe a year down the line, we can find out if they still are, if you've been successful. Hopefully they're not.

Speaker B: No. Thank you for your time, uh, and was pleasure being here and looking forward for the other conversation we can have in the future.

Speaker A: Brilliant. Thanks again.

Speaker B: Thank you.