Bob Laliberte: Welcome back everyone. I'm here with Chief AI Officer Bob Friday. Bob, welcome.
Bob Friday: Thank you, bob, great to be here.
Bob Laliberte: So, we're talking about self-driving networks. I'm wondering for our audience and all the decision-makers, business decision-makers that are out there can you start by explaining what's the journey to self-driving networks and why it's a game changer for modern networks?
Bob Friday: Yeah, you know Bob, for me personally, the journey started when I was at Cisco, and I was trying to convince a large retailer to put this customer experience onto their shopper's phones and what they told me was they were not going to put anything on that network until I could promise my controllers were going to stop crashing, I could innovate faster than once or twice a year and, more importantly, that I could basically guarantee that their customers were going to have a great experience on that network.
And that's when I realized there was really a paradigm shift from, in addition to having to keep AP switches in the network all up and running green, it was really about monitoring the client-to-cloud experience and that's really what I call the self-driving network. That's the paradigm shift.
Bob Laliberte: Got it, got it. That makes a lot of sense. So, let's talk about the journey a little bit more. You know going through the different stages from data to fully self-driving networks. Can you walk us through all those five stages, which I believe are data, insight, recommendation, assisted and ultimately self-driving, and how does each stage build upon the previous one to establish trust and more confidence in the overall system?
Bob Friday: Yeah, so, for those that are like me or a winemaker, what you know is great wine starts with great grapes. Right, and the reason we ended up building an access point at Mist is I wanted to make sure that we can get the right data to answer the question of is that client user having a great experience? And so that's the first step on the journey is making sure you can get the right data to answer the question.
The next part of the journey was really around an organizational shift. That was making sure that, hey, I can get the right data back to the cloud and the people who know that is your customer support team. And so, organizationally at Mist we basically tied together the customer support team with our data science team, because ultimately, that support team is a proxy for your real customer, and that is getting the right data back to the cloud.
The next step in the journey was really getting that data into a format that we could apply AI and math to. And the interesting thing, what we found is for a lot of our customers that had a ton of value, because if the signal is big enough, you don't need fancy math, the average IT person if you give them the data in the right format, they can figure out what was going on.
But where you do need fancy math, that is where we get into AI recommendations, that's where we're helping the IT teams actually sort out those what I call needle in the haystack problems. And finally, the next step in the journey is taking the action of let's actually apply action to fixing a problem. And so those are the five stages of that journey to kind of a full self-driving network.
Bob Laliberte: Okay, that that makes a lot of sense and to build upon each other you need to have the right data in order to get the visibility, etc. You need to be able to tie all that through to make the appropriate recommendations for organizations and then ultimately getting to that point of self-driving almost right, that fully autonomous.
Bob Friday: Yeah, I mean, I give you an example a customer. Actually, when we started with the Zoom and Teams, we had customers having this Zoom Teams problem. So, in there we actually end up building a complete model that can now predict Zoom Teams user experience every minute.
And once we got that up and running, what we quickly found out is what was happening was people were coming to their India site and they were getting rerouted to Australia, and so it was kind of a mixture of different problems going on when we had a misconfigured VPN gateway that was causing pain to a certain set of customers. That is where Fancy Math actually helps IT fight needle in a haystack problems.
Bob Laliberte: Excellent, excellent. That sounds good. Now I know when I hear the term self-driving network, obviously there's an analogy we made to all term self-driving network. Obviously there's an analogy we made to all the self-driving cars that we're starting to see here. So how would you compare the progression that's going on toward to get to network autonomy with the stages that we're seeing in self-driving cars?
Bob Friday: Yeah, we talked a little bit about for the show, about how many of you have actually been in a self-driving Waymo, Uber, right, that is happening today. I didn't think I was going to live this long enough to actually see it, but yes, there are self-driving cars out there like in a sci-fi movie, right, the networking's on the same journey. You know, getting in that self-driving car was a level of trust when I decided to get in that car.
We're in that phase of the journey, inside of the networking phase, where our enterprise customers know that AI is going to be relevant to their business. We are here to help them out now figure out the difference between marketing AI and real AI. And that's that same trust journey. We're going to get to the point where those self-driving cars are going to have a better safety rector than human-driven cars. And we're soon the same journey on the networking side where you're going to trust your network AI assistance on par with your actual human.
Bob Laliberte: Yeah, and I would even argue that maybe the self-driving network is even more important given the fact that the network is getting so much more complex today, with AI being brought in, everything being highly distributed, modern applications. So, all the things that the operations teams need to deal with, being able to leverage AI is becoming critically important for them to optimize performance and ensure security.
Bob Friday: Definitely is their circle of life, because you look at autonomous vehicles. That is going to require a highly reliable network for all the autonomous vehicle use cases we see coming down the pipe.
Bob Laliberte: Yeah, absolutely, and one that is able to give that real time, that performance and so forth. I know a big piece of AI is not just the technology, but it's also the culture. So, I'm wondering if you could talk a little about, as these networks become more autonomous, what type of paradigm and cultural shifts do you see happening with IT teams and organizations, and how do these changes affect the way the teams are managing their network?
Bob Friday: Yeah, I mean, I think, a good point. You know at Mist, when I first started at Mist, in addition to having to build this real-time day-two pipeline that could process all this data, the other big thing was an organizational change where I had to tie my support team up with my data science team and, culturally, IT teams have been trained to use CLIs.
Here we're trying to get them to start to trust their AI system to work with them. So, I think the same cultural change I work with at Mist is going on inside the IT teams. Our customers are having right. They have to learn how to trust this AI. They have to get to know it and trust it like a new hire.
Bob Laliberte: Got it. Yeah, no, that makes a lot of sense, and the hard part is some of this is just going to take time. They need to, and that's why a lot of when we're talking with organizations and they're asking about AI and AI ops, it's well, get started. You need to start using it now. You don't have to turn it on full self-driving to start. You know, if you think about from a car analogy, a lot of times it was just the lane shift, detecting whether there were cars there, right, bringing you on that progression to where eventually be like, okay, well, I trust that it knows where the cars are.
I trust it knows where to keep me in the lane. Now, I'm just going to trust it to keep going. So, you know, being able to leverage that technology to accelerate the time to comfort with the technology. All right, there's a lot of different key technologies that are driving this transformation. I'm wondering if you could highlight some of the most important ones and then maybe briefly explain what their role is in enabling each stage of that self-driving network.
Bob Friday: Yeah. So, this is an interesting question because I have a lot of customers who basically ask me, I was doing AI 34 years ago. When did this all go from research to practical reality? And usually I tell them if you look what Google Trends tells you, it happened somewhere in 2014. That's when we started to see all the searches for deep learning and machine learning.
And when you look what happened over that period of time, it was lower cost, computing storage started, bringing it more relevant and, more than anything else, we see ChatGPT, computer vision and self-driving cars, and that's really differentiating around big models being trained on tons of data.
We've been doing machine learning for a long time and that's a key part of all these solutions, but the really disruptive part of this were these big models being trained on a large set of data, and at Mist, probably our best example is this large experience model. We're training a very large model with tons of data to actually start to predict Zoom, Teams, video collaboration, experience. So those are the type of things that are really bringing AI to reality.
Bob Laliberte: Yeah, no, that makes a lot of sense, and the good part is what you're doing is you're leveraging all of the data specific to your network and the maybe one lily pad over from the network in order to provide those feedback.
And I think you're right about the cultural shift and that somewhat of the consumerization of AI has really helped drive that into the network operations teams to be able to okay, I understand it, it's not a big scary thing, it's not magic, right the last thing you want to hear if you're working in operations. So, we bought a new tool and it's magic, right. It's got to be based in science and repeatable and be explainable.
Bob Friday: And I think what we're hearing from our customers most definitely is they want to know how Marvis got its answer, and so explainability is part of that story. Its in trust. Yes, you need to explain how you got to your recommendation.
Bob Laliberte: One of the major challenges we hear about is the need to build trust in among network teams. How do you see that journey towards a fully self-driving network, helping to build that trust over time?
Bob Friday: Yeah, I think what we're seeing customers right now is, like you say, they're starting with these production POCs. They're starting to understand, where Marvis and Cloud AIOps can actually help them with what we call assisted driving. Right, they're not letting Marvis start to turn knobs or anything, but they are now allowing Marvis to basically find problems for them. So that is the first step to the journey of trust. How often did Marvis actually help them find something?
I think where we are in the journey with most of our customers right now, they're just starting to trust Marvis, to start to let Marvis start to take actions, especially on the data plane. Probably a good example is like radio resource management. Most customers have given their solution full channel power that's done and we're at the next phase of let me touch your switch port, let me reconfigure your missing VLAN. That is the next step we're on right now in the trust of letting AI assistants touch their network.
Bob Laliberte: Excellent. Well, it makes sense at least again. It's all of that get started right, doing something, getting started on that journey is probably the most important thing.
Bob Friday: And I would say the analogy I use to tell people it's like you're new hired If you talk to most IT teams. Like your new hire, if you talk to most IT teams, they don't trust anyone to touch that network. Until you've earned their trust, you're not touching anything on their network.
Bob Laliberte: Absolutely, so it makes sense for the technology you would have to go through the same level of scrutiny as well. Looking ahead, what are some of the key outcomes and benefits businesses can expect as they progress through these stages towards a fully autonomous network?
Bob Friday: I mean, I think the obvious one is, when you talk to our customers who are using Cloud AIOps now, it's definitely fewer support tickets and quicker resolution time. I think, if you look beyond that, if you look at what's happening, like in one of my large educational customers right, once he got Cloud AIOps and running, what he realized he was not always in a reactive mode.
If you look at most IT teams, they basically come in every day and it's a very reactive. Once he got up and running, what he realized is that he had more time to do other things, and so I think that's the other thing we're seeing culturally is, once people adopt cloud AI ops, it lets their IT team start to expand to other responsibilities.
Bob Laliberte: Yeah, absolutely, and I've heard the same thing as well. And when we've done research, one of the common responses that we get is that it's allowed us to again get out of that firefighting mode and go and work on more of the strategic initiatives things that are really fun for the network engineers to really sink their teeth into and be able to develop to enable the business to do more and accomplish their business goals. So, finally, while the fully self-driving stage remains somewhat of a vision for now, what should our audience keep in mind as they prepare for this future, and what steps can they take today that will help them start on this journey?
Bob Friday: Well, I mean, when you tell most customers, I think you're in the business here, I think you guys, all the PowerPoint start to look the same, right. So, the first step of the journey is trying to read out what I call real AI for marketing AI.
And that's usually what I call production POC. Yeah, the best thing to do is pick an actual production site and get an actual cloud AIops solution up and running so you kind of get the value, start to feel the value experience and you'll basically open up your eyes to what's going on in your network. And that's usually what we find with most of our customers who start the journey is they just realize they have more visibility and observability and then you start to work on the trust. Okay, what am I going to trust this solution to actually start touching in my network?
Bob Laliberte: Yeah, no, that makes a lot of sense and I know from speaking to your customers. There's definitely that phase where organizations, when they first put it in and get it deployed, they're finding issues that they didn't even know they had, that weren't even being reported. So, there's that ability to be more proactive in finding those small problems before they come large and create an outage and so forth. And then you know to me that next step of once they get into that it's all about where do they establish that level of trust and enable it to do more and more and take the recommendations, be able to do some more automation as well.
So I think it's great to be able to see how these organizations are getting started on this journey, putting it into the production POC, being able to test, find things out, be able to learn from that and be able to move forward and drive honestly and ends up being a lot more operational efficiency right and enabling them to find and fix problems faster and then be able to again get out of that firefighting mode and into more of a proactive response for the network. All right, great, bob. Thanks, that was great. That was really awesome. Any final thoughts you want to share?
Bob Friday: Yeah, Bob, for the audience, for those of you interested in learning a little bit more on self-driving networks, feel free to join the Q&AI Bob Friday talk show.
.jpg)
