What's Up with Tech?
Tech Transformation with Evan Kirstel: A podcast exploring the latest trends and innovations in the tech industry, and how businesses can leverage them for growth, diving into the world of B2B, discussing strategies, trends, and sharing insights from industry leaders!
With over three decades in telecom and IT, I've mastered the art of transforming social media into a dynamic platform for audience engagement, community building, and establishing thought leadership. My approach isn't about personal brand promotion but about delivering educational and informative content to cultivate a sustainable, long-term business presence. I am the leading content creator in areas like Enterprise AI, UCaaS, CPaaS, CCaaS, Cloud, Telecom, 5G and more!
What's Up with Tech?
Cisco’s Incubation Engine For Agentic AI And Quantum Networking
Interested in being a guest? Email us at admin@evankirstel.com
What if the most human-feeling teammates on your projects moved at machine speed and collaborated across every vendor you use? We sit down with Vijoy Pandey is GM and Senior Vice President of Outshift to unpack how agentic AI and quantum networking move from shiny demos to dependable, scaled systems that actually ship results.
We start by reframing innovation horizons around risk instead of time. Technology risk is where quantum lives today, with foundational challenges from hardware to protocol. Market risk is where agentic AI thrives, as teams test real enterprise use cases beyond coding assistants. Platform risk is the mandate for a company like Cisco: open, interoperable architectures instead of brittle point tools. That lens shapes everything OutShift builds.
You’ll hear a clear blueprint for the Internet of Agents: discover the right capabilities across vendors, grant task and transaction-based access that expires fast, enable real-time many-to-many messaging across voice and video, and close the loop with rigorous measurement that feeds reputation and policy. We spotlight Agency (AGNTCY), the Linux Foundation–hosted open source project backed by a broad industry coalition, designed to make multi-agent collaboration portable and trustworthy.
The proof shows up in production. A healthcare triage flow uses a Webex voice agent that coordinates with insurance and diagnostics agents to verify claims and route patients to the right human expert faster. In networking, a multi-agent pipeline built with Swisscom stress-tests configuration changes like a chess engine, exploring outcomes, generating tests, and catching cascading failures before they hit production. On the horizon, quantum networking becomes a scale-out fabric that links quantum computers, data centers, and sensing devices—accelerating timelines by five to ten years while delivering near-term classical benefits in security and precision.
If you care about enterprise AI, open standards, and the road to quantum-ready infrastructure, this conversation maps the path from uncertainty to impact. Subscribe, share with a colleague who’s building with agents, and leave a review with the one capability you want most from an Internet of Agents.
Can't keep up with AI? We've got you. Everyday AI helps you keep up and get ahead.
Listen on: Apple Podcasts Spotify
More at https://linktr.ee/EvanKirstel
Hey everybody, really excited for this chat today with OutShift by Cisco, the innovation engine of reimagining how big companies build what's next from AI to quantum networking and beyond. VJoy, how are you?
SPEAKER_00:Doing good. It's a pleasure to be here, Evan.
SPEAKER_01:Well, thanks for being here. Really intrigued by your team and your mission within Cisco. What is OutShift by Cisco, first of all, for those who may not be familiar with your team?
SPEAKER_00:So OutShift by Cisco is the internal incubation engine for Cisco. The way we describe our group is that innovation happens everywhere. So innovation happens in the product groups, it happens in CX, in every aspect of business at Cisco, but we drive incubation. And the way we think about this is can we build new products in spaces that are adjacent to Cisco's core businesses? So we think about Cisco, we think about networking security, observability, collaboration, those are the pillars of the business at Cisco. And so we are looking at adjacencies, both in terms of personas as well as in terms of technologies. And that's where we go. We look at everything from ideation to customer traction. And we work with the BUs as that happens.
SPEAKER_01:Fantastic. Well, of course, Cisco knows many things about innovation. But are you trying to rethink the traditional models of Cisco's corporate innovation?
SPEAKER_00:So the innovation is everywhere. So innovation goes into every product, but on the incubation side, I think that's what you were referring to. I mean, we've done Cisco has done incubations in various ways in the past. I mean, we've all been familiar with this whole spin-off, spin-in model that used to happen. I actually come from Google and they have this Area 120 approach, which is another way of doing incubations within the company where people take 20% of their time and people are supposed to work 100%. So that's where the 120 comes from. So this 20 above and beyond that 100%. So there have been various models that we've seen, Cisco has seen. But we wanted to learn from all of those and see what worked, what didn't. And we ended up with this incubation model without shift, where we took a lot of those learnings and tried to make it right. And we've been around for six plus years now, which means that the model is working. That's the way I look at this.
SPEAKER_01:Fantastic. And what's driving Cisco to invest so heavily in these frontier tech areas like a gentic AI, quantum networking? What's the big idea there?
SPEAKER_00:The way we look about incubation is uh, at least the way I think about my job, our job, is to help Cisco reduce risk when it tries to enter a market or a product. And if you think about the horizon models that McKinsey had laid out a couple of years ago, a couple of decades ago, in fact, they talked about horizon three, horizon two, horizon one, and all of those horizons were based on time. So horizon one was right now, within the next year, Horizon 2 was roughly three years out, and horizon three was roughly five years out. Now, as you and I know, Evan, I mean, in this day and age, time is the material. I mean, things are already late, things were late yesterday. And so, in this kind of uh fast-moving paced world, time is not a good metric. So we looked at that framework and we said, how do we categorize risk? And how do we reduce that risk? So we looked at again horizon three, two, one. Horizon three for us is technology risk. Is the technology risky, but it still holds promise? And so one of the spaces that we mentioned, quantum, to us falls in that space where quantum holds a lot of promise, but there are some really hard problems that still need to be solved, and that's technology risk, and we are definitely playing in that space, and we're trying to reduce risk for Cisco in that space. Now, Horizon 2 is market risk. So the technology works, but the market and the use cases may not be stable enough, may not be large enough. The market is shifting quite rapidly, it's uh it's not growing. Sometimes it grows, sometimes it shrinks, sometimes stable. And we have not proven out all of the use cases yet. So there's a lot of potential, but the use cases are still being figured out. And I would say agentic AI is in that bucket where the technology exists, there's a lot of promise, it's been proven out for software development and consumer interactions, but for a whole lot of other areas where there's a ton of promise, we're still proving it out. So that's market risk for us, and it's evolving rapidly as well. And then the third horizon, which is horizon one, so we're going backwards, but horizon one is platform risk, which is as a company like Cisco, we want to play in a platform-based approach. And Getu Patel, our CPO, keeps talking about the platform advantage that companies have. And a company like Cisco cannot fight these point battles. So we need the platform play. And that's purely on the product BU side. So for us, we play in Horizon 3 and Horizon 2, and we work with the BUs to make sure that it becomes Horizon 1 as we move along. So quantum technologies and agentic AI are the two big spaces that we concentrate on.
SPEAKER_01:Fantastic approach. And speaking of big ideas and initiatives, the Internet of Agents uh sounds fascinating. Can you maybe explain, though, what it means to you and why it matters so much?
SPEAKER_00:Everybody's fascinated by agents. I mean, this is this is the talk of the town. And if you think about agents, they feel and behave like humans. I mean, the whole reason for the Turing test to exist was to make sure that AI can behave like humans and be not distinguishable from a human behind a screen. So the fact that we are there, we have silently passed some sort of a Turing test in the past two years tells you that agents behave like humans. But there's a big difference. The difference is that even though they behave like humans and they have human-like characteristics, they are operating on machine speed and scale. So, whereas, Evan, you and I, of course, we are humans, but we cannot be running at whatever 50 gigahertz a second or whatever the number. But these things are running on machine speed at scale and at cloud scale, and even cloud scale numbers are growing every year. So these things are running really, really big and fast. So when you think of humans with machine speed and scale, but still just like humans, they all need to come together in a team, collaborate to solve for bigger business outcomes. So when we started thinking about the Internet of Agents, we basically thought of that as the collaboration platform where you could discover agents and their capabilities, bring them together, and make them operate in a team to solve for bigger business outcomes. Because it's not just one agent from one vendor on one cloud with one persona that's going to solve your business problems. There's going to be a collection of these things that come together and solve a business need. So the Internet of Agents was thought of to handle that problem statement of collaboration between agents and agents, agent-to-agent collaboration, but do it in an open and interoperable manner. So we wanted to make the Internet of Agents open source and interoperable right from the get-go. And we looked at four pieces to that problem. We looked at discovery of agents and their capabilities, the identity and access control and access management of those agents, then bringing them together and helping them communicate and message with each other. And then finally measuring them and making sure that they are being evaluated and they're actually solving for the problems that we all hope that they're solving for. So those are the four phases that internal agents actually solve for.
SPEAKER_01:Amazing. And if anyone can make this work, it's of course Cisco. And what does it take to make agentic AI work reliably in production, in service, at scale, not just in the lab or in proof of concepts and other sort of vehicles?
SPEAKER_00:I'll give you a few flavors of this. So, one, of course, you need great infrastructure. And at Cisco, we are pushing AI infrastructure that is efficient, that can be deployed on-prem, that can be deployed within cloud providers. So we're working with cloud providers as well. And just uh yesterday, we even announced at our partner summit this Cisco AI edge device that can be leveraged for inferencing at scale. So that's the infrastructure layer. But once you've built agents that sit on this infrastructure, you again go back to the problem I just described, which is let's say I'm an enterprise and let's take Cisco as an enterprise. There are agents coming from Cisco, there are agents coming from Microsoft, there are agents coming from Salesforce and ServiceNow and Workday. And a typical enterprise like Cisco deals with all of these agents. Some are security agents, some are workflow agents like ServiceNow, some are HR agents like Workday, some are business and sales agents like from ServiceNow. Sorry, Salesforce. And to set up something simple like a sales funnel, you're probably pulling all of these agents together, making them collaborate, and setting that thing up. Something that used to take a few clicks on a GUI maybe five years ago. Now you're dealing with NLP and with all of these myriad agents coming from various vendors and sitting on different clouds. So the problems here are multifaceted. First and foremost, there's a problem of discovery, which is where can I get the best security agent for my use case? And I don't want to search for something from a vendor, but I want to search from a capability perspective, from a reputation perspective. How can I do those searches across multi-vendor environments? So that's step one. Now once you've done that, you need to give it access and authorization. And role-based, traditional role-based access controls actually don't work because of the problem that we just described earlier, which is role-based access controls was built for a human-led world. But when these agents are shifting personalities, today they are VJoy, tomorrow they are Evan. You can't do role-based access control and you can't give them access for hours or days. So we've taken that problem right from the basic principles point of view and said we really need to look at things like task-based. What tasks are these agents trying to accomplish? Or what tools are they trying to access? Or what transactions are these five agents trying to accomplish? So this is another flavor of how identity changes and access control changes. Then you go into messaging, and you need all of these agents to communicate with each other's either with each other in a many-to-many way because they're all in a team. But they need to be real-time and interactive in nature because you're dealing with video and voice, networking company. I mean, we have to solve this problem. And then finally, with especially with Splunk and what we're doing with them, uh, we are making sure that once all of this is up and running and working, how do you measure these things? How do you how are you evaluating their performance? And once you evaluate them, you go back to ground zero, which is can I can that evaluation feed into discovery as reputation? Can that evaluation feed back into identity and make sure that you're giving the right access? So is there's a closed loop system here that we are trying to solve. And we are we are at the beginning stages of agentic workflows and organizations. But what we feel is that so we are running into identity and discovery problems. But once you start deploying these things at scale, we realize that these things are a real pain point.
SPEAKER_01:Amazing, incredible work. And you mentioned you collaborate with your internal colleagues widely from you know, WebEx, uh the business unit, three Splunk and others. Do you have any examples of how that collaboration looks and works internally and externally with customers and partners?
SPEAKER_00:So all of this that we just described, the Internet of Agents, and actually there is a open source manifestation of Internet of Agents, and that's called Agency. It's spelled AGN T C Y. And uh you can go to agency.org and see the code and documentation. Please come and look at it, contribute. It's it's uh it's a community-led uh organization. It's sits with the Linux Foundation, and it's uh backed up by 80 plus companies, including us, Google, Red Hat, Dell, and Oracle as the foundation. But all of this is great. But what is it solving? What is it there for? And so, to your question, what are the use cases? And so we've been working with internal teams, we've been working with design partners and customers on shipping some use cases on this Internet of Agents stack. So I'll give you two examples, but there are many more. Uh, one is in the healthcare space. And so let's say I have a claims that I want to go through, and uh uh this is in the healthcare space, and I'm talking to a well-known hospital, and the hospital is trying to figure out which department to send me to. Uh, how big is this issue? Is it something that needs uh I should pay for it or it needs to go to an insurance provider? So all of these things need to happen. So there's a little bit of a pre-diagnostics or pre-uh debugging that needs to happen before I'm sent to a human. So, what has happened is that we've built a multi-agent tech workflow where I, as a customer of this hospital, come to a WebEx interface. And this WebEx interface is actually again voice-based and video-based. So it's using all of the messaging that we just described. And we I start interacting with this uh voice agent, and the voice agent in the background is talking to an insurance claims agent and it's talking to a health diagnostics agent to make sure that VJoy is valid and he's not just taking the day off because he couldn't complete his homework or something.
unknown:Right?
SPEAKER_00:So that's happening in the background, and then once there's reasonable sort of through the evaluation mechanism, once there's reasonable certainty that this is legit, then I'll be handed off to a human, maybe a doctor or somebody else, that can actually go ahead and take my take my call. So that's one use case. The other use case, which is near and dear to our heart in the in the networking side of the equation, is we've built this entire pipeline for network config validation. Now it sounds pretty trivial and simple, but the way I describe it is like you're playing chess against deep blue. And think of it as you are trying to configure any infrastructure. In this case, it happens to be the network, and we work with Swisscom, a big telco provider. But let's say you're configuring any infrastructure, even your laptop. Now, when you make a change to the infrastructure, in this case a network, even a small change to a particular switch or a router can have some ramifications somewhere else in the network, like a butterfly effect. And as humans, it's very hard for us to comprehend a small change here making an impact somewhere else. And we make this more complicated, like cloud infrastructure and things like that, it's even becomes even more difficult to comprehend. But for an agent or a set of agents exploring this entire space like Deeplue can, it can look at every chess move out there possible. So what if we build something like that that can explore every configuration change problem out there through this simple change and then generate tests for those changes and test my change against all of those possibilities, then whatever I'm trying to do is guaranteed to work and not cause a problem somewhere else. And that's this multi-agent workflow that we built for network config validation with SwissCom, but we are rolling it out through again our products into many, many other areas as well.
SPEAKER_01:Fantastic. And speaking of networking, quantum networking is one of your key focus areas. Uh, what are you trying to achieve there?
SPEAKER_00:Okay, if you think about Cisco, uh networking company, and if you think about paradigm shifts that happen in the industry, we've seen one paradigm shift happen through a genetic AI and genetic AI. And it was relatively easy to catch up to this paradigm shift in Genitive AI because it's still a small matter of programming or it's just software. So it's easy to catch up, and as long as you're passionate and aggressive as an organization or as an individual, you can catch up to what's happening. The next paradigm shift that we are on the cusp of is quantum computing. And that one is rewriting computer science theory over the past 70 years. I mean, you you're going to rewrite, it's going to change the way CS is taught in schools. So it's like really fundamental in nature. And while that happens, it's going to be, if that happens, like in a stop function, which we believe it will, it's going to be very difficult for organizations to just catch up through software because it's a fundamentally different thing. So as Cisco, being the networking and security player, at least that we are, of course, as observability and lab, we want to build out the quantum internet, which means we want to build out the network that connects quantum computers, so a quantum network fabric. Then we want to build the quantum network that connects quantum data centers together. And then we want to build a quantum network that connects quantum sensing devices together. So we want to build the entire hardware protocol software stack to enable quantum networking at scale. And that's what we are trying to achieve. But as we do that, what we are also figuring out is those same building blocks of a quantum network can be utilized today in classical use cases, in classical computing, to solve for problems that were really hard for classical computers to solve, but become really easy for a quantum network to solve. So it's a it's a duality of purpose for us in quantum networking where we're solving for the scale-out problem, the networking problem in quantum, but that same network and the same building blocks are actually solving classical problems today. And that's what we're after.
SPEAKER_01:Fantastic. And what's the uh potential benefit of a quantum internet to customers, uh, what industries might benefit first?
SPEAKER_00:So if you're thinking about the quantum compute aspect of it, there are lots of benefits of going towards quantum computing in general. So, first of all, I mean, so if you think about drug discovery, if you think about uh material discovery, if you think about logistics problems, all of these are really hard problems. And Feynman uh had a pretty interesting quote which said that nature is quantum, so let's stop simulating nature using classical rules. And that's what quantum computing is trying to do, and that's what we're trying to approximate using these really large foundation models, which becomes really, really hard because now you're building these really massive data centers to simulate a problem that might not, should not take that much space and computing power. And so the promise of quantum computing is that with relatively small compute environments, you can explore simulation spaces that are very, very large and solve these really hard simulation nature-based problems, maybe the understanding of the universe. And so, through the network, what we are trying to do is accelerate the timeline for achieving those goals by five to ten years. Because if you just went and did scale up, which is make bigger and bigger quantum compute nodes, you're looking at maybe five years, ten years, twenty years, depending upon who you talk to. But if you were to collapse it and say, let's put the principles of cloud and distribute systems in place and scale out, connect a bunch of these things together into a distribute computing environment through a network that we bring to the table, then you suddenly are reducing that timeline by five to ten years. And that's where we see a lot of excitement.
SPEAKER_01:Incredible opportunity. Can't wait to see how this unfolds. And um, just a final question on the team. Maybe give us a peek behind the curtain at uh AlShift by Cisco. You must be very mission-oriented. It must be really exciting and also fun to work on these projects.
SPEAKER_00:So the team is amazing. The team is actually looking at the entire pipeline, just like a bunch of startups mode, from ideation all the way to customer attraction and success. So we are composed of innovators and passionate people, passionate about emerging tech, that are looking at ideas, that are building products, engineers, product managers, uh people who market this stuff and are passionate about uh making the messaging simple and educating the community, uh, people who are biz def people who are hunting for design partners, and then people who are customer support folks who make sure that we provide white glove treatment to our first initial design partners because they are the most important ones and they'll make our technology live or die. So we are actually based, we have a we constitute all of these functions. We are small, nimble, and we try to bend the rules and make things happen.
SPEAKER_01:Well, thanks so much for uh the vision and sharing the mission. Really exciting. I can't wait to see progress over the months and years ahead. Appreciate your time for joining.
SPEAKER_00:Thank you, Evan. It was great to be here.
SPEAKER_01:And thanks everyone for listening, watching, and checking out our TV show, techimpact.tv, now on Fox Business and Bluebird. Thanks, everyone. Thanks for joining.