Juniper Al-Native Network Platform: Momentum and New Developments

0:00 first of all mobility field day is an

0:02 extremely important event for us at

0:05 juniper mist it's actually very personal

0:08 to us anytime and every time we have

0:10 come out with something which we feel is

0:12 good we've always presented here first

0:15 so today you'll hear from bob and

0:17 sudhira and my team on what we are

0:19 unleashing as we enter into 2025 and

0:23 2026 so let's just legal disclaimer you

0:26 all can read that but here is the main

0:28 thing as we continue to march on our

0:32 journey to unleash the network of the

0:34 next decade we have continued to

0:36 experience durable momentum in the

0:38 industry now i'm sure you'll say "hey

0:40 this is mfd this is not a financial

0:42 presentation." but i'm an engineer i

0:44 always start with the outcome and then

0:46 tell you how i think this has been

0:47 happening campus and branch revenue

0:50 great growth orders continue to just set

0:54 records but here's the main part what is

0:57 the winning formula we've always said we

1:00 are fundamentally focused on enriching

1:02 end user experiences while ensuring that

1:05 you get the lowest trouble tickets when

1:08 you enable a juniper miss network this

1:11 slide hasn't changed you've seen this if

1:12 you've been at any of my presentations i

1:15 always talk about this this is a day in

1:17 the life of suj whether i'm on teams

1:19 whether i'm on zoom whether i'm at some

1:22 business critical application on voice

1:24 or collaboration we get this fix

1:26 pixelization we get this frozen thing

1:29 two years back we used to say you're

1:31 muted now we say you're frozen expecting

1:34 someone on the other end will know what

1:36 we're doing so as you know juniper miss

1:38 philosophy has always been up is not the

1:42 same as good which means a network being

1:44 up does not mean you're getting an

1:46 amazing enduser experience and it gets

1:49 worse as you start looking at headless

1:52 devices you know humans if you have a

1:55 problem you can complain to someone if

1:57 you want to robots when they have a

1:59 problem they just stop

2:01 working so how do we deliver on this

2:04 philosophy it's very simple you've heard

2:06 this before this is the mist ai native

2:09 networking platform it's all about the

2:11 right data which bob is going to talk

2:13 about more importantly the right

2:16 infrastructure we are today i would say

2:20 still the only vendor where for our

2:23 customers christmas comes every

2:25 wednesday because every thursday there's

2:28 a global production push across all the

2:30 cloud instances worldwide

2:34 last but not the least you get the data

2:38 you have the right infrastructure to

2:39 handle the data but the right responses

2:41 become extremely important bob will talk

2:44 about aidriven support how we are the

2:46 only ones in the industry who drink our

2:49 own champagne i prefer that to eat your

2:51 own dog food okay and simple thing when

2:54 a customer has an issue on a juniper

2:56 miss network the first person to answer

2:58 that ticket is marvis in the end before

3:01 i hand it over to sudir and bob this is

3:03 what we will say this is my personal

3:06 commitment to each and every customer or

3:09 prospect who's considering juniper mist

3:12 we will show you the fastest roll out

3:14 the fewest trouble tickets and

3:16 fundamentally drive business outcomes

3:19 how do we do all this i'd like to

3:21 welcome to stage bob friday because as

3:24 we release these new innovations no one

3:26 better than a bob and sud show to get us

3:29 excited about that thank you bob please

3:31 come over yep thank you suj yeah so good

3:34 morning everyone as suj says mobility

3:37 field day is the highlight of the year

3:38 for myself and the miss team because it

3:41 is where we do introduce our latest and

3:43 greatest innovations for the year and

3:45 this year we have some very cool things

3:47 to show you now for those who know me

3:50 you know i make a barrel of wine every

3:52 year i always say great wine starts with

3:54 great data great ai starts with great

3:57 data great wine and grapes i guess close

4:01 enough but anyway when susan started

4:04 mist we did not build an access point

4:06 because we thought the industry needed

4:08 another access point we built that

4:10 access point because we wanted to make

4:12 sure we could get the right data to

4:14 predict and optimize the client to cloud

4:18 experience now since joining juniper in

4:21 2019 what you have seen us do over the

4:23 last 6 years is extend that marva's data

4:26 pipeline ingestion from the client to

4:29 the access point to the switch to the

4:32 sdwan router to zoom and teams data and

4:36 last year we announced marvis minis that

4:40 was the industry's first synthetic user

4:42 that let us make sure that all critical

4:45 network services and critical business

4:47 applications were up and running before

4:50 that network open in the morning now

4:52 what you're going to be seeing us

4:54 introduce and talk about today is marvis

4:57 minis client to cloud

4:59 applications and marvis minis sol that's

5:03 going to be the industry's first

5:05 synthetic user slee in the market now

5:10 once you have all that right data in the

5:13 cloud you have to organize it in a way

5:16 so we can apply simple cadere math or

5:19 fancy data science math to get to the

5:22 root cause now last year i would say

5:25 mist is still the only vendor out there

5:28 that has moved from a paradigm of

5:31 managing just the network elements to

5:33 actually managing the client to cloud

5:36 user

5:37 experience when we collect that data we

5:40 collect data for every user minute for

5:42 every user on that network now last year

5:46 you saw us introduce the zoom teams

5:49 large experience model this allowed us

5:52 to take zoom teams data join it with

5:54 network feature data build models that

5:56 could actually predict that zoom user

5:58 experience this year you're going to see

6:00 us announce the generalized zoom teams

6:04 model this is going to allow us to help

6:06 customers who don't have zoom teams

6:08 using user on network

6:10 understand video collaboration user

6:12 experience on their network now when it

6:15 comes to data science we all have access

6:18 to the same data science algorithms

6:20 we're all using the same pytorch

6:23 tensorflow libraries you know what makes

6:26 us different and miss is really we have

6:30 basically joined the customer support

6:32 team to the hip of our data science team

6:35 that is that aidriven customer support

6:37 we talk about you know when you look at

6:40 customer support they're the ultimate

6:42 proxy for our customer the fewer

6:46 customers the fewer support tickets that

6:48 our support team sees is the fewer

6:50 tickets that our customers are sending

6:52 us so we are still the only vendor that

6:55 is using its own cloud ai ops solution

6:57 in its customer support team and finally

7:01 conversational interface now i have been

7:03 a big believer of natural language

7:05 interfaces since i started the company

7:08 you know i believe natural language

7:09 interface is going to be the next user

7:12 interface into the network you've seen

7:14 us go from cli to dashboards natural

7:17 natural language interface is that next

7:19 thing last year you saw us introduce

7:22 genai public doc search we are now

7:26 integrating genai technology into marvis

7:29 to give marvis a voice and making it

7:31 easier for our customers to get answers

7:33 from junior public docs this year what

7:36 we're going to be announcing is genai

7:39 customer support we're going to start to

7:42 help our customers actually get access

7:44 to their network data and an ask their

7:46 network

7:47 questions and then finally marvis

7:50 actions this is a self-driving component

7:53 of marvis you know and what you're going

7:56 to be seeing sadier and i talk about is

7:59 a marvis actions facelift we're going to

8:02 be taking marvis actions from assisted

8:05 self-driving to full self-driving this

8:08 year

8:12 and we've all heard about ai agents this

8:13 is going to be the next big thing that

8:15 we're adding to our data science

8:19 toolbox now i think many of how many of

8:23 you have done the self-driving whimo

8:26 user experience

8:29 self-driving how many of you were at

8:31 wopc in phoenix this year i know most of

8:34 you were it was freaking amazing right

8:36 to watch these self-driving cars pull up

8:38 at that airport felt like you were in a

8:40 sci-fi movie now what that demonstrates

8:44 is we have the technology now to build

8:46 some pretty amazing self-driving

8:50 experiences now i don't think our

8:52 enterprise it customers are quite ready

8:54 to hand over the keys of their network

8:56 to marvis but i will tell you miss

8:59 marvis is the farthest along this

9:02 self-driving path marvis does a great

9:05 job right now of assisting it people to

9:08 find problems in network and we're

9:10 starting to see enterprise it customers

9:13 start to trust marvis to actually fix

9:16 things like stuck ports and missing

9:18 vance we are on the journey to

9:24 self-driving now i always tell people ai

9:27 is a concept is the next step in the

9:30 evolution of automation it is not a

9:33 single model or algorithm per se when

9:35 you look under the hood of that

9:37 self-driving uber you will find a lot of

9:40 technology there to get that

9:42 self-driving experience to work when you

9:44 look under the hood of marvis you will

9:46 find a lot of technologies to get that

9:48 self-driving networking experience to

9:50 work machine learning has been around

9:52 for years what is really disrupting the

9:55 industry has been these deep learning

9:57 models where we're training very large

9:59 models on very large data sets and what

10:02 we're starting to see is the gent ai is

10:05 getting added to that data science

10:07 toolbox and that is going to be the next

10:09 big element in getting us to that

10:11 self-driving networking experience

10:15 now when susan and i started miss we

10:18 knew that we were going to have to build

10:20 a whole new microservices cloud

10:22 architecture from scratch to do real

10:24 time day2 operations what we quickly

10:27 realized after we got the company up and

10:29 running that we were going to have to

10:31 build a new

10:32 organization you know this is where we

10:34 actually tied our data science team to

10:37 the support teams you know that is back

10:40 to that customer support is the proxy

10:42 for our customers right you know we have

10:46 to get make them happy which makes our

10:48 customers happy now if you look what

10:51 we've done since 2018 the data science

10:54 team and support team have got together

10:57 every week surely narrows are here they

11:00 basically work with our support team on

11:02 a weekly basis over the last seven years

11:05 what you see here is we have gotten

11:08 marvin

11:09 consistently to an efficacy around 80%

11:12 plus now when you look at the chart you

11:14 see it's kind of flatline that is

11:16 because it's an ongoing basis we're

11:19 continuously adding new network features

11:21 and we're seeing new challenges in the

11:23 network marvis efficacy is human

11:27 resources human reinforcement learning

11:30 everything we learn in that customer

11:32 support process goes back into our

11:34 product that's why marvis is

11:37 consistently getting better year on year

11:41 bob i have i have a question here do you

11:44 think we'll ever get to a point where

11:45 marvis will be you know we'll have the

11:47 answer like 95 or 100% of the time i do

11:51 okay and we'll talk about it i think

11:52 with a combination of marvis minis large

11:55 experience model and this new enentic ai

11:59 technology i think we're going to bring

12:01 marvis to that next 90% plus level now

12:04 how far do you think away we are from

12:05 that

12:07 come back next year okay we'll be here

12:11 pivoting on uh that uh do you see

12:15 harvest uh spearheading organizationwide

12:19 automation

12:20 initiatives or in addition is it

12:22 aligning with seeing more enterprise

12:25 adopting organizationwide automation

12:29 yeah it's a good question i think you

12:31 know for those who are following these

12:33 agentic ai

12:35 mcp i see uh agentic ai becoming a new

12:40 nonlinear nondeterministic programming

12:42 language i see mcp becoming the next in

12:46 agent friendly

12:49 api and so i see organizationally you

12:52 know similar to how organizations build

12:54 scripts on top of apis we're going to

12:56 see enterprises start building agent

12:58 frameworks on top of mcps