2932: Innovating for Tomorrow: How Outshift by Cisco Delivers Next-Gen Solutions

[00:00:01] How is artificial intelligence reshaping the landscape of technology? And what new developments can we expect from a leader in the industry? Well today I'm recording at Cisco Live where I'm joined by the senior vice president of

[00:00:17] Outshift by Cisco. His name is VJoy and Outshift is at the forefront of Cisco's innovation, especially in the realms of AI cloud native applications and quantum technologies. So I've invited him to join me today and share his insights on

[00:00:33] the latest AI innovations are not only advancing at Cisco's product line but also setting new benchmarks for customer experience and most importantly some would say technological responsibility. So buckle up and hold on tight as I beam

[00:00:49] your ears all the way to the show floor here at Cisco Live in Vegas where you can sit down with myself and VJoy. So a massive warm welcome to the show. Can you tell everyone listening a little about who you are and what you do?

[00:01:03] Absolutely, first of all pleasure to be here. I'm VJoy, I run this group called Outshift by Cisco. It's the incubation engine for Cisco so we build products that take Cisco into new markets especially towards new personas both

[00:01:18] user personas as well as new buyer personas. We've been structured like a bunch of startups. So think of it like an internal Y Combinator where we have of course product building capabilities but we also have go-to-market muscle. So everything from engineering, product, marketing, customer success, sales, they

[00:01:36] all sit within the organization because building the product is the easy part taking it to market, having user empathy, having the right messaging, right customer success motions is the hard part. So we've been structured that way

[00:01:51] and that's what we've been doing for the past I'll say three and a half years. Incredibly cool, you must love your work it's a real exciting part to be in and before you came on the podcast today and sat here with me I was doing a little

[00:02:03] research and I was reading that Outshift's mission is to build what's next in emerging tech for Cisco. So I'm not sure how much you're gonna be able to share here but can you share some recent AI innovations that have come

[00:02:15] out of Outshift and how they're transforming customer experiences? And the reason I ask that is there's a lot of hype around AI and machine learning but it's great to hear about the problems that you're solving and the real

[00:02:26] product at the end of it. Yeah I mean actually we've been using AI for a long time, a whole bunch of products but what's happened as we all know as consumers as well since November 2022 generative AI has been democratized even

[00:02:42] though generative AI has existed since actually I would say generative AI has existed since the 1960s when we had the first chatbot called Elisa that was built over a rule engine but the technologies that power generative AI

[00:02:57] have actually moved and transformed and they've gone from statistical rules to statistical engines to deep neural networks to now neural networks with context which is what transformers are which is what the new wave of generative

[00:03:11] AI looks and feels like. So I would say that has driven a lot of pressure on organizations to do something with it because it's so easy to use and there's so much immediate value that you get out of generative AI that not only are we

[00:03:27] using it in our daily lives as consumers but organizations are looking at this and saying what can I do with it especially when it can generate content it can summarize content, it can transform content from one shape to

[00:03:40] another and it can do all of this through a multi-modal environment so you're looking at text to video and video images I mean it's like it's all over those modes of content creation and consumption that generative AI is

[00:03:52] getting good at. So there's a lot of pressure on organizations to use it what we found out in our recent Cisco AI Readiness Index was that 97% of organizations really really really want to build something using generative AI

[00:04:07] but only 14% are prepared to do anything with it. So there's a massive gap and so one of the things that we at Cisco are looking at is how do we bridge that gap and what are the challenges? So you think about it foundation models, so LLMs

[00:04:24] their capabilities are actually exponentially increasing over time. You get one new model and one new capability every day and so you you get GPT-4 O, you got Sora, you've got all of these things so that's going really

[00:04:39] swell but then on the organization and the enterprise side consumption is a massive challenge because what we think of four main reasons. Number one people don't even know where to start. Is there an easy start button? Is there starter

[00:04:53] code? Is there a sandbox? I can start playing around with my ideas on what my use cases could be and then I can maybe narrow down the use cases that I need to build. So that's number one. Number two, once I do that and I'm comfortable with my

[00:05:07] use cases I need to customize it to my context. So I need to bring my data, my knowledge bases, my policies, the models that I've acquired or have passed some responsible AI framework that exists within the company. So I need to bring

[00:05:24] all of those things together to my use case. So customization is the second big problem. I have my use case, I've customized it. The third problem is am I getting the return on investment because these things are not cheap. So you're paying

[00:05:38] an arm and a leg to your favorite model provider. Are you better off? Are you that much better off the amount that you paid for it? So ROI analysis, cost analysis, topic analysis, workflow analysis. Are you getting efficient over time? Is this the

[00:05:54] right model to pick versus the other one? That's the third problem. Then finally when you're all said and done and you're ready to deploy your application, can you deploy that scale with trust, safety, and security? So those are the four problems

[00:06:06] that we see organizations struggle with. And so we released this product called Modific. So Modific.AI actually and it allows customers to walk through that journey. Starting with easy start, let me experiment, figure out what use cases

[00:06:22] to make sense, all the way out towards deploying it in production and being it safe, secure, and trustworthy. Incredibly cool. And for anybody that's not here at Cisco Live, me and you have been living and breathing it over the last few days.

[00:06:38] But for anyone that's not been here, how would you say Cisco is integrating AI into some of its existing products and services to ultimately enhance its capabilities and address the evolving needs of your customers that you just

[00:06:50] mentioned? Yeah, so I think the way I would frame that is, so Cisco thinks about AI in sort of two main buckets. First is productivity and then product. How can we use AI? And when I say AI, it's primarily generative AI for now. But

[00:07:08] think of AI in all shapes and forms and machine learning in all shapes and forms. But productivity is how can every employee within Cisco be more productive? And can generative AI actually help these employees be more productive?

[00:07:24] So you're looking at use cases for developers, right? So can you use things like GitHub Copilot or ForgHR? Can you have build systems that do resume summarization and categorization? Can you look at marketing? Can you summarize

[00:07:39] blog posts and make it a linked post? So there's a whole bunch of these use cases that we're looking at that improve the productivity of every employee within Cisco. So that's one category. The other is product. How can we use AI in the

[00:07:53] product space? And even there, we break it down into two categories. One is AI to improve our products, existing products. And then what products are we building to improve AI itself? So in the first category, AI to improve products, you see

[00:08:09] things like assistants. I mean, there's the 5-1 assistant that Jitu talks about. Yes. There's the summarization and catch-me-up features that exist within WebEx. We have a product called Panoptica that has a pretty nifty feature which says,

[00:08:23] it's like a lens. We call it a lens. It's like a Google lens, but it's not really a camera that is being pointed. But you can take your cursor and point it to anywhere in the UI and it will tell you exactly what that entity within that UI

[00:08:35] means and does. So these are all features that use AI to improve the experience of existing products and security, observability, networking, and collaboration. But the most interesting bucket is actually building products that move the bar forward when it comes to AI. So Jonathan and I were saying, the

[00:08:58] Fabric, the partnership that we're doing with NVIDIA, that is to build out training and inferencing clusters for AI. So that's products to improve AI. Similarly, Jitu talked about security for AI. That's improving how AI is consumed and improved upon. Motific falls in that category as well because it

[00:09:18] allows organizations to consume generative AI and actually build applications that pertain to your use case. So that's the more interesting bit and there's a lot happening and I'm just giving you a simple feel of this, but yeah, there's a lot

[00:09:33] happening. And there are so many businesses wanting to do AI right now, but not knowing where to start. That is the thing that we'll be talking about more and more. You mentioned that powerful stat a few moments ago. So what

[00:09:47] are the big challenges that business leaders are coming to you and asking for your help with? Are there any trends in the type of conversations that you're having? Yeah, that's actually great. I mean, it's been a good learning exercise

[00:10:00] for us since we announced Motific back in February at Cisco Live Amsterdam. And even before that, as we've been talking to customers and even internal teams, like I mentioned earlier, we are going after this set of developers that the

[00:10:18] industry knows as citizen developers. So these are developers who are non-technical people. So if you have a SaaS product that is using AI, those guys are good because they're technical people, they understand the nitty gritty, they can

[00:10:34] use AI, they can build out the features, all good. It's the HR teams, the sales teams, the finance teams, the legal teams. I mean, these are the teams. They don't have developers, deep tech developers, but they want to leverage the power of

[00:10:51] generative AI to solve some of these pain points that we were talking about. And like I said, they don't even know where to start. And so the one learning experience for us has been, we want to enable those citizen developers, because

[00:11:04] of all the stats that we came across, and this is from a survey that GitHub did, is that there are 27 million users, developers in 2024. That number is going to go up to 45 million developers in 2030. The need is going to be 150 to 500

[00:11:24] million. So there's a massive gap in the need of developers. And we believe the only way to solve it is to enable all of these use cases by citizen developers, through citizen developers using generative AI. Because now you just need to know

[00:11:43] natural language to generate a use case, to build out a use case. And that's what I think Motific is trying to do, is enable that journey. And what we've found out is people need a place to experiment, to then figure out what use cases make sense,

[00:12:00] and then go on that journey. And one of the most interesting pieces here is talking to Liz Sertori, who's running our Customer Success organization. She has realized that this journey is something that Cisco can help our customers with. So she's launching AI advisory services

[00:12:19] that actually will allow customers to go on this journey all the way from, again, what use cases make sense all the way to deploy it in production, to if you're really serious,

[00:12:30] and you want to build your own custom models, let's allow you to do that as well. So I think that to me has been the base learning where citizen developers are the future to leverage the power of generative AI for a whole variety of use cases.

[00:12:45] And from the conversations that I've been having here with people from Cisco, one of the things I picked up quite quickly is you're very proud of this commitment to responsible AI as well.

[00:12:56] So what steps are you taking to ensure ethical AI practices in your developments? And how do you balance innovation and responsibility? I'm sure it's a question you get asked a lot. So the way, yeah, that's actually a great question. I think if you

[00:13:11] think about our purpose statement is to power an inclusive future for all. We have been pretty passionate about that purpose statement. We've been pretty passionate about human rights and privacy, about data governance, and about data privacy issues right from the get-go.

[00:13:30] And AI only accelerates that journey. So we've had this responsible AI framework even before generative AI became democratized back in 2002. Even to deal with what you would call predictive or traditional AI or machine learning, there may be various terms people use.

[00:13:50] But we've been shipping AI within like WebEx or networking or security all this time with that responsible AI framework in place. And it goes through their entire lifecycle around is your data catalog? Do you know where it's coming from? Do you know what purpose is it

[00:14:08] going to be used for? So that we are very careful about what goes out and what comes in. And we also are very careful about if we can use synthetic data, we prefer to use synthetic data

[00:14:20] to train our models and to make sure that we are effective. So we're pretty rigid about that. We're pretty passionate about that framework. So both myself and Jen Yokoyama, who sits in the legal team, she's our deputy general counsel, we co-chair this responsible AI committee

[00:14:40] that actually goes through this assessment, this framework, this process for every product that we ship, but also for every productivity use case that we use internally. Because it's both in and out. We've been evolving that. It's an evolution. We've been evolving that over time

[00:14:59] as generative AI committed, it got this first evolution. But to your question around that friction, it's a great question because I think of responsible AI in the same way as we think about security. If you have a piece of software that is insecure, in my mind,

[00:15:18] that piece of software is unavailable. It doesn't exist. Because you would not use an insecure piece of software. Similarly, if you have a piece of software using AI that has not gone through

[00:15:29] responsible AI practices, to me, that piece of software does not exist because you would not want to use it. And so the entire framework of how we think about responsible AI, we are embedding it

[00:15:41] into the same development practices that exist for security. So to us, it's one and the same. So it needs to be part of the process, part of the pipeline in the same way that we think about

[00:15:52] security. So security and responsible AI, we need to pass through all of these checks before we release anything. And when we're talking of friction, there's also been a lot of talk around AI and the big tech layoff in the industry. You were talking about a shortfall of developers

[00:16:08] ironically as well a moment ago. So how do you think this narrative is going to evolve? Yeah, I mean, I think, and I'm not going to name names here, but there are some people who are

[00:16:17] pretty well known, who are actually doing, I would think disservice to the community by spreading these notions that AI is going to be taking away jobs or is going to automate things in a way that

[00:16:33] humans won't be responsible anymore. I'll take a simple example of developing code. And I think one of the things that we keep hearing is, yeah, developers are going to just disappear. I'm just taking that example, but you heard that statement in so many other verticals as well.

[00:16:48] Just a few days ago, I heard the same thing about math. It was like, yeah, we won't need math majors. I'm like, really? And so to me, it's like, yes, AI will get better and better. And there is a

[00:17:01] transition that's going to happen from being assistive in nature, so assistants, to having agents that actually do tasks for you. But I think the way we think about all this is that it's just

[00:17:15] raising the level of abstraction at which humans need. So we were doing, I mean, we don't, if you're building a house, you don't create bricks anymore. You use, I mean, at least humans don't create bricks anymore. They get created somewhere else that is all automated and you

[00:17:34] lay the bricks, you build your infrastructure, you're done. We just need to go higher up the stack and that's where I think humans will end up being. But the subject matter expertise, whether it's co-development, whether it's legal, whether it's finance, whether it's HR, that still stays

[00:17:50] with the humans. And I think there's a little bit of that push and pull happening, the layoffs are happening. That's primarily because people are like organisations are shifting resources towards AI because of the pressure in that hype cycle that's happening right now. But things will

[00:18:06] settle back to normal. And bringing it back to your work and everything that you're working on here, how do you outshift leverage AI to drive advancements in other emerging areas like, let's say for example, cloud native applications, edge computing and quantum technologies? A lot of

[00:18:22] excitement around that too. Yeah. So I think this goes back to the framework that I described, which is AI for product and product for AI. So we are leveraging AI more and more to build

[00:18:35] features into products that were more difficult to do with humans or almost impossible to do with humans because of the fire hose of data and information that these platforms gather.

[00:18:47] So anyway, you see this deluge of data coming in is easier. And so every time you have that problem, you leverage AI to build some of those features out. So some examples, of course, as an assistant,

[00:19:00] we've talked about the lens. Those things have existed in the product. We have this notion called an attack path where we are thinking about how can an attacker come in and compromise your modern application? And that's part of this product that we built around securing your modern

[00:19:18] distributed application. And so we have this notion of an attack path. And so far, we had a team called a red team. And their job was to constantly hack into cloud environments and modern applications to actually surface vulnerabilities that can then

[00:19:37] be solved through Panoptica. And you can then build these attack paths that customers can then fix. But now we're looking at generative AI to actually create those attack paths in a more

[00:19:48] automated manner. So the red team is actually now a gen AI red team. So that's a pretty nifty feature that we've invented. But the exciting bit, I would say to your question on quantum is like

[00:20:02] really, really exciting. And it's just going to change the game dramatically for all of us. So we're spending a ton of time figuring out what it means to do quantum networking and quantum security because that's our sweet spot. We will not be building

[00:20:17] quantum compute nodes. That's not where we play. But once you have a bunch of these quantum compute nodes, how do you connect them into a LAN-like construct or a data center-like construct? How do you connect these data centers across wires? And so that's where we are focusing

[00:20:33] on the quantum side. And then there's a pretty palpable security problem that we're going through right now, which is called store now and harvest later or decrypt later attacks that governments and FinTech and these very highly regulated and sensitive industries

[00:20:49] are pretty paranoid around. So we are tackling security for those kinds of use cases as well. So quantum networking, quantum security is something that we are spending a lot of time on.

[00:20:59] The other thing that we are pretty excited about is this whole journey from AI assistants to AI agents. And we feel that the future is going to be a bunch of these agents working together

[00:21:12] to solve a business problem for somebody, whether it's a team, an individual, an enterprise. How do we enable that world where a bunch of these agents come together and solve a business outcome from somebody? So I'll just tease about that and leave it. But

[00:21:30] it's a pretty exciting time. That's a wonderful teaser. And if I was to nudge you a little bit further and look further ahead, what are some of the most exciting AI trends or technologies or converging technologies that you think will significantly

[00:21:43] impact the industry? And how do you think Cisco will position itself to lead in some of these areas as well? Yeah, I would say the agentic frameworks that I just described, like agents

[00:21:53] coming together. And right now what we're seeing is things like chatGPT or if you're looking at Anthropic or some of these other foundation models, they are performing what's known as one-shot approaches to solving a problem. They have these big mammoth models that you go and

[00:22:12] ask something, that's a prompt. You get back a response, whether it's code, whether it's Q&A, whether it's summarization, but it's a one-shot approach. You ask something, it's one big model,

[00:22:23] it does something for you. Think about the number of times a human would be successful in that kind of approach. Would you go to a singular person to ask questions around, where should I go on

[00:22:36] vacation? What is it that I'm suffering with in my heart because I had chest pain? And can you give me financial advice? There's not a singular human that can answer all those three questions at the

[00:22:48] same time. So a one-shot approach is not the right approach. And so what people are looking at is what's called a multi-shot approach, where you basically build subject matter experts. And think of it just like humans, and these people need to come together,

[00:23:02] solve a business problem. And I think this is a distributed agent AI problem. And if Cisco is good at something, it's good at enabling distributed computing because that's what we've done right from the get-go. We've connected computers, we've connected software, we've

[00:23:22] secured distributed environments, and that's what we plan to do for AI as well. And you're someone that's had hundreds of conversations during Cisco Live. You've been on stage, off stage, on the show floor. I'm curious, when you get that plane ride home,

[00:23:36] what are you going to be thinking about? What excites you about everything that you've seen and heard? Yeah, the one thing that has been amazing is when we've been talking about some of these problem statements and the approaches that we have, every customer,

[00:23:51] I don't say this lightly, but typically when we go and talk about a problem statement, you'd get maybe a 70% hit rate or 80% hit rate. And customers will be like, some will be like, yes, I'll buy this. Some will be like, show me more. Some, yeah, come and talk

[00:24:06] to me in maybe two years time. When you talk about AI and some of the challenges that we just went through, every customer has been 100% hit rate. They're like, do you have it today? I'm

[00:24:18] ready to pay money and deploy it today. So that's been a pretty exciting bit for us. Love that. Well, thank you so much for joining me today. And for anyone listening who just wants

[00:24:27] to keep up to speed with some of the developments here at Cisco Live and beyond, is there anywhere you'd like to point everyone listening just to find out more information? Sure. I mean, of course, you can look at ciscolive.com, which is the website for this

[00:24:39] event itself. But Outshift is outshift.com. And you can look at Motific, you can look at Panoptica, all of the research that we're doing in quantum, the open source projects that we have,

[00:24:51] all of the next gen, Verizon, Verizon two and three projects that we're working on is all available at outshift.com. We've covered so much into 25 minutes today. I do urge anybody listening to check out that

[00:25:05] information. But more than anything, I know how busy you are as well. Just thank you for dropping by and talking to me today. Thank you for having me. So as we wrap up today's episode, I'd love to pose a question to every single person listening.

[00:25:16] What role do you see AI playing in your field over the next five years? And also what ethical considerations should companies prioritize as they integrate more AI into their operations? As always, this is where I invite you to join the discussion share your views. This is a

[00:25:33] dialogue, not a monologue. So email me tech blog writer outlook.com x Instagram LinkedIn, just at Neil C Hughes, let me know your thoughts. And we'll keep this conversation going. But thanks for tuning in today. Make sure you catch the next episode of tech talks daily for

[00:25:49] more insights from leaders who are helping to shape our digital future. But thank you for listening today. And until next time, don't be a stranger