Summary
Change Management & Resistance
Agent Building
Token Costs & Monetization
Data Quality
Data Ownership
Notes
Transcript
So from my journey perspective, I'm born and brought up in India. My dad's in the army. So every three years, we were part of a new school.
And of course, more than any schooling, it was just-She's seeing cities every three years, building new relationships, so that's where a lot of the dam may adapt.
You'll see through my career, I have done multiple rules. And I think a lot of it is related to the army background. Then last, after high school, I went into engineering, I did environmental engineering. So nowhere close to what I'm doing currently, but I'm going to be joining the United Nations Environmental Program, and that will be the light of the day. Environmental engineering came to Purdue, where I did my master's in environmental engineering, and then had a few offers, but I happened to chance to meet with this guy who was would come there for an info session.
And in the spirit of giving directions, we ended up making an offer for the startup.
The company was based out of Atlanta. It was in point of sale and payments. And we were just 200 people when I started with them. It was called Radiant Systems.
And a lot of work, the first 10 years of my career, I'm very focused on just the same side of the house.
Account management for customers like British Petroleum, for customers like Push Start, for big customers. And as that happened too, came in, spent 10 years, the company got acquired. I went to NCR and NCR was National Cash Register. They also have evolved over time. They now drive it from banking perspective, all of the ATMs of the world. NCR runs those. And as that has gone through, I came here to the Bay Area.
I came to the Bay Area to work for Palo Alto Networks. And Palo Alto Networks is in cyber security, so a lot of retail and FinTech background, and then moved into cyber security. Palo Alto Network was a two billion dollar company when I started. We were less than 6,000 people. and over the last five years, Apollo also really set the norm on scaling. And we did multiple acquisitions during that time frame.
And all the acquisitions were integrated and got Palo Alto from a Gartner perspective. Whether we talk about SASE, whether we talk about Firewall, a lot of progress on Palo Alto Networks. And currently, we're over $8 billion when I left, with around 15,000 of those people from the state. Thank you. Recently started with Cisco and I am running the product strategy and operations for a gentleman named DJ Sampa.
He is a founder, CEO, came into Cisco from a company called Armor Blocks. and now really helping at the helm of transforming how Cisco is driving and how Cisco is investing in the AI centerhouse.
Oh yeah, he was supposed to be our speaker. The other guy was supposed to be our speaker. I put one slide in. This could have actually started here, but it's done for. By Sandy and Leonard, and the whole premise of Cisco's company was for me to be connected to computers and the network became the most powerful thing. So as we have evolved over the years, whether it is telecom, mobile, internet, networks becoming powerful and more powerful.
And as we think about the AI revolution, we are estimating thatThe next book in mind will be pre-pied for the next few years. So we are focusing a lot on ensuring that the network is in good place in Los Francisco.
Product front end operations, I will take a few minutes to talk about what is, so when you, when I'm gonna start up, product front end ops start at the center of What is policy and engineering building and what are we taking to the market? And as the You can put a lot of things in there. You can put comms in there. You can put financials in there. So in a small company, it's anything that does not fit into your salary.
-Cisco has so many domains, so many different views. How do we sort of bring everything together? The role that we have, the parts that we are bringing to market are cutting across all of the social needs. We are bringing all of the business together. So a lot of my job sits at the intersection of ensuring that we are building the right products, we have a plan to sell it, but also connecting the cross-social org that we have at Cisco. A lot of time The market is also in question.
Every domain. -Now that we have been here, together is how are we having one motion-pointing motion across the scale?
We have so much data in the system. We are putting that data across various domains and creating our own models. So instead of running--... We absolutely can still run this. But using our own data, because we have leveraged it, so one, the cost is lower. Second, the performance is higher. The latency of--What do you mean? We just hear you.
So you can choose, like how you can toggle in strategy also. You are able to choose which model you want to use, whether you want to thinking model, reasoning model.
So, Cisco Cloud Control is a platform that we launched it for the availability in June of this year, and we are going into today as we go into June. -And so, as I said, it's bringing various, we actually, every domain had their own sort of version of a dashboard. All of that things are bringing together under the self-control of. What is that, Rowan?
Previously, we used to store the multiple sign-ons that we were doing. So we have also said, we're going to do one sign-on, and everything comes under-after you sign on, you're able to see all your assets in one place. All your topology in one place. You're able to say, tell me, in Atlanta or in New York, what are all my. You're able to see in what condition.
Is an AI, AI, we call it AI Canvas, that is able to look through all of that, you can, AI Canvas is there, we're leveraging the AI models, where you can ask questions and go back and forth.
You're also not just asking questions and understand what the issue is, but also we are now saying you're able to remediate.
You're able to say, OK, change the policy. The agent will tell you, hey, I can do this and what needs to be changed. And you're able to change the policy and things like that and go ahead and remediate it right there. Thank you. And that's what I would say, so if you look on the left, we have with the AI says, you'll need my network health summary across campus, where I'm telling you at the data center, you're able to get all of what is happening and look at the recommendation and what needs to be done.
You have customized workflows across multi-vendors.
So another thing we are focusing on is an open ecosystem. You are able to bring various other vendors in.
So we have recently partnered with 50 plus of them.
We are working with them and we can bring their agents in.
And from a cloud-controlled studio perspective, you can have the agent builder where the customers can build their own agents. or the apps within the ecosystem and work with one time.
Yes, so what we have done is, there will be certain general, I will just tell you this, with the network demand that is coming,A lot of folks will have to upgrade in the hardware side. It's nothing to do with Cisco, that's that.
But in general, like when video was coming, there was a whole rush around upgrading your infrastructure. And it's gonna be the same here, because if the demand is gonna go too high in the next few years, If you are on very old systems, you will have to move away.
But the point I might be keeping is not to remind-Connecting, you log in, if you're using some of our products, you're able to log in and we are, whether it's Meraki, whether it's security, you're able to bring it in. Okay, that is all I had. That is all I had on the Cisco side. I didn't want to spend too much time giving you a whole overview of Cisco, but I wanted to tell you what we are working on. This is very transformative for the company.
So that is the, the paradigm has completely shifted with the AI coming into space. What if you had asked us what I showed you about Agent Studio, Two months back, three months back, there was zero, nothing on this planet. This all happened in a span of two months.
And I think that is what is, I feel like, things that have worn from the last, I would say, 10 years now. And we have to be very adaptive towards that. And that's what we're trying to say. When we talk to the customers, now customers don't ask us 12 months of roadmap. They used to ask us 12 months of roadmap. Because they know in two months, if I'm going to put a roadmap out there,It's Wednesday. Things will evolve.
We built AI defense, which was a security product. All 95% of it was built using AI tools. Spectrum and development. put it through the execution and it was delivered and it is now GA. So right now I feel from a product management standpoint, it's less about execution.
It is more about what needs to be built.
That is where the value is coming from. Execution is becoming--Easier and cheaper.
When we were building on the AI campus, we had someone, we spent six months building AI cameras, and then one of the leaders, one of the technical guys said, you know what, we spent a weekend, probably five days, and built a version of it which is better than what we spent on building it. Also because there were new tools that were released. between those six ones. We have to dance and focus on police and we have to completely throw away all the six months of work that we have done.
I feel a lot stuck in the car. Like at the CPO level, so one Cisco didn't have a CPO.
We've got everything together under one leadership. So now let's go ahead and start off with this.
And the guidance from the top is also set that everyone, we are reviewing every few weeks what is the air adoption in the woods. There are a lot of opus there. We know there was a time when Cisco fell, but we lost some of the, we were not on the journey, but Cisco had gone down in the cloud.
We missed the cloud. So we know if Cisco double, double down on the AI revolution, it's not going to succeed. So there is so much focus. This is it. Every breath.
I don't see a startup within Cisco. It was, we started with, we had started with 200 people, now we are on, Thank you. So very small org compared to Cisco. The number of employees at Cisco is around 80,000. So if you think about the org that we are in, it's a startup driving, it's at the helm driving the revolution here.
But what we're also doing is, we have a channel that says AI mission. And we are putting everything on there. There's nothing hidden behind the behind the door that we are making.
We are bringing in the teams, the leadership, when I say leadership, team directors and above. We're bringing them in our discussions on that channel. I'm making the decisions.
So what is it that you want to be famous for in the new world, accepting that you're going through this transformation process? I would say jumping cops. Gen-tech ops. Gen-tech ops, yes.
Gen-tech ops, okay. And when I say gen-tech ops, it is actually bringing a lot of those together. How do you, so right now, how do you--But when you talk about probability, when you talk about visibility, All of those in our genetic thought story Security is part of it. Observability is part of it. Figuring the visibility into your network and hardware is part of it.
That is the, how do you have one place? to view one advocate, to understand what issues are happening, to mitigate the issues,And so that, I think that could be the real.
The second thing I put on there was Because leadership needs to decide that innovation is important. When some of our leaders expect a lot of stability, people don't take the risks.
So having that, ensuring that the computers are openAnd that is where I tell the team to hold back if the leadership is not-you have to respond.
This does not mean agreeing to everything. This does not mean, it just means that you have a space, you're thinking differently, you're working with your teams as they are bringing ideas to the table. And again, I would say that it encourages You're awesome, Keith. Courage and forward motion. With that, I wanted to open upFor this conversation and what, I would love to hear what challenges you are facing in your-Thanks.
At least. In my view There's a lot of resistance and change which is... All this because the whole, I think,News about jobs and placing jobs and everything that's involved with this. So it's more of a culture issue. More about leadership and asking people to experiment and try and do new things and encourage that. What has worked is that we do sessions where we share best practices. from different I oversee different companies. So from one company to the next. So like, where are the adopters? Why are the serious people get excited, get the other people excited and try new things?
But still it hasn't been handed to most people. So like my biggest challenge, I think it's our ability to change and then-The people who are usually are more set in their ways. It's probably harder for them to come.
And you people, the more-so that whole dynamic is the biggest issue. I feel like with a lot of the AI, it is leveled the space, right? They have wrong experience and they've gone, "Hey, I have experience. "I know how to run this." But these young kids have. They know exactly and they're fearless. So a lot of what we also doing is having our leaders, it's very difficult to run an AI org if our leaders are not hands on.
You have to start--I just want to know, for example, And I said, I don't want to build a slide of onboarding.
I don't want to onboard you based on your. We've got an onboarding agent. You know, it's not that, it's easy. It's just a first step. And you have to be like, okay, let me just go in. People talking about it, just get in.
So that, I think, it's a takeaway, as you guys all know, for people who did not raise your hand. Figure out whatever agent you want to build, but build something. One of the things we are doing is, like, between the leadership team, we are doing a .mp, basically all of our groups that happen on a daily basis. Special needs? I'm not the CEO of the store, and the details are just a little. What we do is--Every day, he sends me a oneof his thing.
The agent basically summarizes his top of mind.
And I do this thing. I will summarize my talk of mine and send it back to him. So we are always alive. It's a quick one minute view of what that is.
--Correct, so what we are also doing, for example, the onboarding agent, now we are putting it in a place, we are talking engineering on-board, they can use the and they can further update it. So we are putting it in a common space for the agents to be available.
What you're realizing is the pricing of tokens. So you start building, using-everyone did a big push of making sure that, hey, we can use and all of that. But understanding the tokens, I don't thinkWhen I look at financials, Month by month we were doubling.
On this section. which is great because we wanted to, but all the pricing was covered. So we are also keeping an understanding of, I think right now it's all across the,Everyone is figuring this last song right now. But it is something that we all need to keep in mind. So do you recommend that we create some sort of tracking dashboard mechanism and look at that like monthly or even more frequently?
I do it a lot. But see, again, our work is small, so monthly is fine. It also depends upon the size of the work.
Who owns the data at the point that you start working with them? Is it combined ownership? Do we still hold on to that?
and how does that value attribution work? And is that a driver for wanting to do this?
So I think we have not made any compensation changes, but we are rewarding people for like, we have competitions, we will have who's using it. Thank you. so they can then go on and betray the rest of the folks. There's championships like we are having where if you are submitting, you're asking people to submit. You've made it and submitted. Tell us how you're using products. Tell us how you're using thought. And you submit it and we are rewarding like, what, maybe F or something.
So nothing crazy on the competition side, but motivating them and encouraging them by recognizing the people who are doing it. That's what we've done, but nothing on the compensation.
Yes, on the KPI, so on the board, so what we're presenting to the board, one of the big things is AI efficiency for the work. So that is one of the pillars that we are talking about. very closely, and that is at the CPU level, we are tracking how we are adopting AI across all the needs of the industries. So it needs to be-needs to be at that table. If it's not at that table, youAny questions?
We have been talking about monetization and what does it mean for us. It's a very timely-because we are releasing our product. What does that mean? So what we are saying is that we thinkWhat you're saying is, if you have Cisco product, it's free. To come out, the base is free. Okay. But when you start using AI Canvas to do remediation, to understand and make a policy change, because those require So someone got to take the cost of the token. So as a customer, either, and we are still figuring out the details because we literally, we had a discussion yesterday on marketization. But we started, the base for us has to be free because we don't want to charge you for bringing the platform together.
So we know that he's the basis free. We are going ahead and letting everyone know it's true. And then as things evolve, tokens, yes, there will be pricing to the tokens, whether you charge or whether you pay directly to the front-end models, there will be charges there. So we've still figuring out that. And then as you bring more value, there will be some compensation. That's how it is really going on.
You mentioned earlier when I was talking to you about tokens. I'm out a little bit. I would think it would probably be a good idea in the token dashboard to not only see how much user-Sure.
-And I think that it is, as we start figuring out, that is another thing, like how do you show customer visibility into the tokens? -Not just the customer level, but who?
And I would say that is why when we say that we have an option to say we want the model that we have created, right?
Because you don't want a large-album model for something small thing that you're doing. So we are building small one model for that reason. So that way you're able to say, okay, you know what, this one will require heavy lifting, to your point, and let's use this, and for this one, it will not require heavy lifting. It's a simple idea. And we are allowing, even if you don't know what to choose for, we'll choose that for you, based on the query or the prompt you've put in.
How do we make sure that we only use it forI'm not going to use it for that reason. I think we should be all in. I am very bullish about it. I think we should be all in and everyone's figuring this out. Yesterday, we were working with another provider for our AI research. We have a whole research farm. So for our AI research, we realized some of the token pricing just went up for the renewal. And it went up 25%.
So now, one, I'm negotiating, or we are negotiating too, but also figuring that out.
So we are able to at least-and we know that data is thin. The data that sits in Cisco in terms of all of our products, and the fighting in that. So that is what we are investing in, and making sure that that data, like unified data, fabric, we find unified data fabric, is across. So that's what we are doing, but--Oh.
Like every org, like right now, the orgs, I'll give you an example. One of the things that's on my, on my charter for next year is called creating an org group.
I'm also working on figuring out, can we have, like, some type of a system or a... that I can just type, find me this and figure things out. So just on my thing for a goal for next year from an ops perspective. And the question then becomes is not every data is gonna be clean. How am I gonna say which data, I've not solved that part of it yet. But for Cisco, what they're doing is they're using unified fabric. That I think is already a clean set of data.
But the thing about AI for us is it starts finding finding signal in areas either because my data was messy and I never saw it or what have you. And so I'm wondering how your, I've used maybe some customers or Cisco is thinking about that question of like, Um... Going about cleaning up the data And this trade-off of like, well, let me just clean up the data worth it. The question is very important to me versus-exploratory, as now this tool is actually telling me where I should be going.
Because there are things that youYou might have a document that you put up a customer to publish it. But in Hanoi, there will be multiple origins of the.
Maybe incorrect and they have more and more time, how do you make sure that you're explaining the right information from the right data? Right now, what we are trying to do is we are marking a lot of things with a certain tag also, but I will say we have not solved that. It's not solved. And it's on the list to figure it out. I don't have an answer on that.