Juniper Al-Native Network Platform: Momentum and New Developments

First of all, mobility Field Day is an extremely important event for us at Juniper Mist. It's actually very personal to us every time and every time we have come out with something which we feel is good, we've always presented here first. So today you'll hear from Bob and Suzie and my team on what we are unleashing as we enter into 2025 and 2026.

So let's just legal disclaimer, you all can read that, but here is the main thing. As we continue to march on our journey to unleash the network of the next decade, we have continued to experience durable momentum in the industry. Now, I'm sure you would say, Hey, this is MFD, this is not a financial presentation, but I'm an engineer.

I always start with the outcome and then tell you how I think this has been happening. Campus and branch revenue, great growth orders continue to just set records. But here's the main part, what is the winning formula?

We've always said we are fundamentally focused on enriching end user experiences while ensuring that you get the lowest trouble tickets when you enable a Juniper missed network. This slide hasn't changed. You've seen this.

If you've been at any of my presentations, I always talk about this. This is a day in a life of seja. Whether I'm on teams, whether I'm on Zoom, whether I'm at some business critical application on voice or collaboration, we get this pixel pixelization, we get this frozen thing.

Two years back we used to say, you're muted. Now we say you're frozen. Expecting someone on the other end will know what we are doing.

So as you know, Juniper, miss philosophy has always been up, is not the same as good, which means a network being up does not mean you're getting an amazing end user experience and it gets worse as you start looking at headless devices. You know humans, if you have a problem, you can complain to someone if you want to. Robots, when they have a problem, they just stop working.

So how do we deliver on this philosophy? It's very simple. You've heard this before.

This is the missed AI native networking platform. It's all about the right data, which Bob is gonna talk about. More importantly, the right infrastructure we are today.

I would say still the only vendor where for our customers, Christmas comes every Wednesday because every Thursday there's a global production push across all the cloud instances worldwide. Last but not the least, you get the data, you have the right infrastructure to handle the data, but the right responses become extremely important. Bob will talk about AI driven support, how we are the only ones in the industry who drink our own champagne.

I prefer that to eat your own dog food. Okay? And simple thing, when a customer has an issue on a Juniper Miss network, the first person to answer that ticket is marvelous in the end.

Before I hand it over to s and Bob, this is what we will say, this is my personal commitment to each and every customer or prospect who's considering Juniper Mist. We will show you the fastest rollout, the fewest trouble tickets, and fundamentally drive business outcomes. How do we do all this?

I'd like to welcome to stage Bob Friday because as we release these new innovations, no one better than a Bob and Susie show to get us excited about that. Thank you. Bob, please come over.

Yep, thank You Sujay. Yeah, so good morning everyone. Sujay says, mobility fill day is the highlight of the year for myself and the mist team because it is where we do introduce our latest and greatest innovations for the year.

And this year we have some very cool things to show you. Now for those who know me, you know I make a barrel of wine every year. I always say great wine starts with great data, great AI starts with great data, great wine and grapes I guess.

Uh, close enough. But anyway, when Susan and I started this, we did not build an access point 'cause we thought the industry needed another access point. We built that access point 'cause we wanted to make sure we could get the right data to predict and optimize the client to cloud experience.

Now since joining Juniper in 2019, what you have seen us do over the last six years is extend that Marvins data pipeline ingestion from the client to the access point to the switch to the SD-WAN router to Zoom and teams data. And last year we announced Marvis Minis. That was the industry's first synthetic user that led us make sure that all critical network services and critical business applications were up and running before that network opened in the morning.

Now where you're gonna be seeing us introduce and talk about today is Marvis Mini's client, the cloud application and Marvis Mini's, SLE. That's gonna be the industry's first synthetic user SLE in the market. Now, once you have all that right data in the cloud, you have to organize it in a way so we can apply simple, severe math or fancy data science math to get to the root cause.

Now, last year, I would say Mist is still the only vendor out there that has moved from a paradigm of managing just the network elements to actually managing the client to cloud user experience. When we collect that data, we collect data for every user minute for every user on that network. Now last year you saw us introduce the Zoom team's large experience model.

This allowed us to take Zoom team's data, join it with network feature data, build models that could actually predict that Zoom user experience. This year you're gonna see us announce the generalized Zoom teams model. This is gonna allow us to help customers who don't have Zoom teams using our network understand the video video collaboration user experience on their network.

Now when it comes to data science, we all have access to the same data science algorithms. We're all using the same PyTorch Tinture flow libraries. You know, what makes us different and missed is really we have basically joined the customer support team to the hip of our data science team.

That is that AI driven customer support. We talk about, you know, when you look at customer support, they're the ultimate proxy for our customer. The fewer customers, the fewer support tickets that our support team sees is the fewer tickets that our customers are sending us.

So we are still the only vendor that is using its own cloud AIOps solution in its customer support team. And finally, conversational interface. Now I have been a big believer of natural language interfaces since I started the company.

You know, I believe natural language interface is going to be the next user interface into the network. You've seen us go from CLI TO dashboards. Natural Natural language interface is that next scene.

Last year you saw us introduce Gen AI public doc search. We are now integrating Gen AI technology into Mars to give Marvis a voice and making it easier for our customers to get answers from Junior Republic Docs. This year what we're gonna be announcing is gen AI customer support.

We're gonna start to help our customers actually get access to their network data and ans ask their network questions. And then finally, Marvis actions. This is a self-driving component of Marvis, you know, and what you're gonna be seeing to Sadir and I talk about is a Marvis actions facelift.

We're gonna be taking Marvis actions from assisted self-driving to full self-driving this year. And we've all heard about AI agents. This is gonna be the next big thing that we're adding to our data science toolbox.

Now I think many of, how many of you have done the self-driving Waymo user experience self-driving? Couple of you. How many of you are at WOPC in Phoenix this year?

I know most of you were, it was freaking amazing, right? To watch these self-driving cars pull up at that airport, felt like you were in a sci-fi movie. Now what that demonstrates is we have the technology now to build some pretty amazing self-driving experiences.

Now, I don't think our enterprise IT customers are quite ready to hand over the keys of their network to Marvis. But I will tell you Miss Marvis is the farthest along this self-driving path. Marvis does a great job right now of assisting IT people to find problems in network and we're starting to see enterprise IT customers start to trust Marvis to actually fix things like stuck ports and missing Vance.

We are on the journey to self-driving. Now I always tell people AI is a concept, is the next step in the evolution of automation. It is not a single model or algorithm per se.

When you look under the hood of that self-driving Uber, you'll find a lot of technology there to get that self-driving experience to work. When you look under the hood of Marvis, you'll find a lot of technologies to get that self-driving networking experience to work. Machine learning has been around for years.

What is really disrupting the industry has been these deep learning models where we're training very large models on very large data sets. And what we're starting to see is the gen AI is getting added to that data science toolbox. And that is gonna be the next big element in getting us to that self-driving networking experience.

Now, when Susan and I started this, we knew that we were gonna have to build a whole new microservices cloud architecture from scratch to do real time day two operations. What we quickly realized after we got the company up and running that we were gonna have to build a new organization. You know, this is where we actually tied our data science team to the support team.

You know, that is back to that customer support is the proxy for our customers, right? You know, we have to get make them happy, which makes our customers happy. Now if you look what we've done since 2018, the data science team and support team have got together every week.

Surely naras are here. They basically work with our support team on a weekly basis. Over the last seven years, what you see here is we have gotten marvis consistently to an efficacy around 80% plus.

Now we look at the chart, you see it's kind of flatline that is because it's an ongoing basis. We're continuously adding new network features and we're seeing new challenges in the network. Marvis efficacy is human resources, human reinforcement learning.

Everything we learn in that customer support process goes back into our product. That's why Marvis is consistently getting better year on year. Well, I have, I have a question here.

You think we'll ever get to a point where Marvis will be, you know, we'll have the answer like 95 or a hundred percent of the time? I do. Okay.

And we'll talk about it. I think with a combination of Marvis minis large experience model in this new Ingen AI technology, I think we're gonna bring Marvis to that next 90% plus level. Now How far you think away we are from that?

Come back next year, we'll be here. Cheer, pivoting on uh, that, uh, do you see harvest, uh, spearheading organization wide automation initiatives? Or in addition, is it aligning with seeing more enterprise adoption, organization wide automation?

Yeah, it's a good question. I think, you know, for those who are follow these agentic ai, MCPI see, uh, agentic AI becoming a new non-linear, non-deterministic programming landing. I see MCP becoming the next in agent friendly API.

And so I see organizationally, you know, similar to how organizations build scripts on top of APIs, we're gonna see enterprises start building agent frameworks on top of MCs.