Cisco's AI Transformation Journey From Fragmented Systems To Smarter Workflows

[00:00:00] A big thank you to Denodo for helping me make more than 60 monthly interviews possible across the Tech Talks network. And as businesses move from Gen AI to Agentic AI, trusted data becomes everything. Everything from Gen AI to Agentic AI, Denodo is helping organizations build intelligent, secure and scalable AI solutions with data access, governance and explainable results.

[00:00:28] So build AI that you can trust and do it with Denodo. And you can learn more by simply visiting denodo.com. What does digital transformation really look like inside one of the largest engineering organizations in the world? Because let's be honest, most transformation stories sound great on a keynote stage.

[00:00:53] But inside large enterprise, the reality is often far messier. Legacy systems, spreadsheet chaos, teams with deeply embedded ways of working and that mantra of we've always done it this way before. And years of technical debt that nobody wants to touch. But today's guest has spent several years doing exactly that inside Cisco. His name's Jason Andrews.

[00:01:19] He's joining me at Team 26 in Anaheim, where he will share how he oversees operations across Cisco networking in a business unit representing more than 22,000 engineers and product managers and roughly 40 billion in annual revenue.

[00:01:36] So today we will talk about Cisco's massive transformation journey from fragmented legacy tooling and 85 plus Jira instances into a unified system of work built around Jira, Confluence, Jira service management and AI embedded workflows.

[00:01:53] And he will share some pretty big stats today and how he sees this as a transformation rather than a migration and how Cisco reduced tooling, tooling, tooling spend by 54 percent. All while dramatically accelerating reporting and operational visibility. Oh yeah, we will discuss everything from scenario planning, technical debt and the reality behind that so-called SaaSpocalypse that we keep seeing in our news feeds.

[00:02:21] So if your organization is wrestling with modernization, AI adoption or large scale operational change, you will find some real world insight in this conversation. But enough from me. Let me introduce you to my guest right now. Thank you for joining me on the podcast today. Can you tell everyone listening a little about who you are, what you do? So Jason Andrews, I'm the vice president of engineering operations for Cisco networking. It's the largest business unit inside Cisco.

[00:02:50] We've got around 22,000 engineers or engineers and product managers kind of bunched together representing about $40 billion revenue per year for Cisco. So I run engineering operations in Cisco. What I do is I end up having, I have program management. So we lead all the programs and projects and releases. We also, I also have the tools function, which again, the Atlassian stack sits there. Any custom develop applications.

[00:03:16] And I also have a function called global lab services, which is about 1.2 million square feet of data center space for our engineering labs. This is where they do regression test beds. They're soldering chips to boards. And we kind of run the overarching service for that. And one of the things I wanted to discuss with you today is I think many people listening, when they think of Cisco, they possibly think about how you help transform other organizations in multiple industries. But you've been on your own transformation journey.

[00:03:42] I mean, you've replaced what decades of legacy tooling inside a company the size of Cisco. Got to be a huge undertaking. So when you first looked at 85 plus Jira instances, thousands of employees with established ways of working already, what convinced you that transformation had to happen? I mean, I think just looking at the way people are working and, you know, going in. And I think I will always go back to a story. Like it was my first couple of weeks at Cisco meeting with a program manager on the team and sat down.

[00:04:08] He goes, yeah, I spent about 10 to 12 hours a week creating a PowerPoint for our meeting every Thursday. I'm like, that's ludicrous. It's way too expensive. As I started digging into why we were doing that, we had disconnected systems and disconnected data. It was very spreadsheet driven. And we had some Excel masters, which is great, right? But it's not really scalable, right? So we really spent a lot of time thinking about how you actually, you build that out. And we had a great forcing function. We had an end of life. We wanted to kind of exit using Rally.

[00:04:36] And so between that, I looked across the environment and I said, obviously, I don't know who wants to do outless defect management. That's not using Atlassian, let me be honest, right? And I just went through it. I kind of used that forcing function in the timeframe to just say, guys, let's build a standard way of work. Let's actually leverage before the term was marketed, right? A system of work and actually get this thing set up so we can actually measure things and see the outcomes without having people spend all their time munging together reporting. And then we actually don't solve any problems.

[00:05:07] And when I was doing a little research on you, I read that you described this as not migration, a transformation. So what was the cultural shift to move? I think it was 15,000 plus people towards a unified system of work. And why do large scale transformation projects often fail when they only focus on the technology? Yeah. And I've talked to so many people here and they're like, yeah, we couldn't, you know, we moved all of our, it's your moving a house, right? You've got, or your body or something. You move all of that to the new house. Like, do you really want to do that?

[00:05:36] Or do you want to go, you know what, do we still need these things? Are these processes still valid? Is this the way we want to work? And I really challenged the team as we sat down and go, you know, I'd sit down with one of the resources and he's, you can't change this. Because that's, and I'm like, why? And he goes, because that's the way we work. And I'm like, why? Because the process. And I'm like, cool. Who owns the process? He goes, you do. I go, feel free to change it. But the question to him was like, you've been over here 25 years. You've, you've seen this ups and downs. How would you do it if you had to start from scratch? I would do this. I'm like, do it.

[00:06:06] This is, we have a one set. When you do these major tool transitions, it's a 10 year investment, right? You're not going to really mess this up for 10 years. It takes way too much effort and coordination. So I'm like, guys, get it. Like really take an opportunity to transform. Don't move all the, your debt and your scar tissue and all these things over. Really focus on how you want to work going forward than the ways you've done for years. Right? I think I've talked to so many leaders here at various companies, different Alassie events.

[00:06:33] And it always amazes me because they, they struggle with their migrations because they're literally lifting everything up from 85 different instances. And I'm like, guys, why, like, why wouldn't you want to start from a green field? Right? Because erase that 15 years of technical debt that you've built up in the system and actually move. And if, because if you don't do it, then once they're in the toothpaste is out of the tube, you'll never, it's harder to come back than later and go, you know what? We don't want to use these fields. Oh, well, you know, we can't change that. It's in place. So I think you have one shot to get it.

[00:07:08] People listening in organizations that have struggled to provide measurable difference, ROI and all those things. But one of the standout numbers that I was reading about you guys was a 54% reduction in tooling spend alongside 40 times faster reporting. So for business leaders listening, where did that real value come from? Is it cost savings alone or, or finally having the operational visibility right across the organization? I mean, when you, when you look at 85 instances, there were 85 server instances in the rally

[00:07:37] instance, we probably had 40 administrators across the org. Now, most of them weren't full-time, part-time, part of person's job. And so we, you're actually freeing up capacity. The organization is you centralize and standardize really does help it scale. I mean, we run with a very, very lean team. I think we have all of eight admins overseeing 15,000 users, right? And convents, Jira, JSM, and that entire stack. And it, it's really about, you know, I think when we lowered that total commerce ownership, we also signed a big EA right with Alassian when we did it.

[00:08:07] So our pricing dropped again, you were paying 85 server bills. Each one of them had a, you know, license. Oh, we need to buy 300 users. How many do you have? We have 250 per growth. When you start munging that into the bigger thing, we lower our licensing costs through working with Alassian. And then as you go through that process, you just see a lot of value getting created because people are always like, oh my gosh, our cloud migration costs more. It's like, not if you do it right, right? If you transform as part of it and actually lift and don't do just a lift and shift, you actually have the opportunity.

[00:08:35] You will see the ROI when it comes to the faster reporting. That's just kind of what I would talk about earlier, right? As you start to get things in a common format, you don't need the spreadsheet. You can literally just query the system. We've moved from PowerPoint to just leveraging, wearing our project team meetings and confluence. The dashboards are live there. Nobody's having to compile them. The same guy I talked to the 10 or 12 hours, he said, I spent about 10 minutes a week prepping for that same meeting today. He's been doing that function of the year. So I always go back about once or twice a year and I'm like, hey, how long has it taken

[00:09:05] you? Now he's down to 10. I probably should have stopped asking because it's like, okay, this is annoying, dude. You're not going to make up that six additional things or six additional minutes. You can have a good point in a podcast, but it really has been a measure of growth of the organization. We've really been able to scale. Like we've, everybody's gone through budget cuts. I don't know what company that doesn't. Right. And as we go through, we've actually, because we've leveraged tools to such an extent, I personally believe we've been able to kind of insulate ourself from 40 to 50% of that. Right.

[00:09:33] As we saw, you know, budget cuts come down, we've actually added services, different functions inside our business, more tools, people, et cetera. But that's been organic growth through the fact that we're not doing manual reporting. So when I have to lose a program manager, like I can actually, it doesn't completely fall over because now that one program manager used to be able to do two can now do four. Right. And then I'm able to take that investment and put it other place to deliver more value to the business. So I think it's, that's building out the system work and now teamwork graph.

[00:10:00] And now at that point, I did not understand AI was going to be a thing. Right. I think it hit us all like a brick, but then as I started looking and I see that, you know, kind of the way it lasted instead, I'm like, wow, this really positioned as well in the air to really leverage it. So I'm actually really excited over the next year of how much we're going to be able to kind of drive, uh, in productivity. Like I see some like really huge gains organizationally for specific job functions, but overall, I know it's, it's, it's really got improvements, but again, that all starts with having a common set of data, right?

[00:10:30] A data flow diagrams and all these things. So you know how your information moves across systems and you have standard ways people are using them. Then you can really leverage it to kind of do the spec driven development and all these things and all this stuff that really, I always tell people like if, because you have the data structure, the things you hate doing every morning will now be done by AI and that's good because now you take that time from things you hated, go to things you love and that you really like doing solving real business problems. And I think that's really got the team excited.

[00:10:57] And many large enterprises often struggle with fragmentation between engineering teams, operations and business and leadership, for example. But how did bringing Jira Confluence, Jira Service Management and Forge together to, uh, how did it bring in all that together, help break down some of those silos in practice? I mean, the first thing you're going to have with any engineering team is everybody thinks they're special. Everybody's right software different than the other team. I've worked in cloud organizations, hardware embedded systems, et cetera.

[00:11:25] They, uh, somebody always writes code, they test code, they release code. And so you don't need all these special workflows. And it took a lot of time convincing that, right. Uh, to really build that, uh, that, that value stream. And it, it was really, and I think some of it was me being a bit of hardheaded is not giving up on it as we went through because they've come to the, no, but we're special. And I'm like, go ahead and get set up. If this doesn't work for you in 30 days, right? I'll, I'll add another workflow for you in the special. And what you quickly realize they go, cool.

[00:11:52] So you mean I get all these automations and all I need to do is drop this one thing and you can automate all. Yes. So we had to show value on the backend, right? So I always say like, if I'm going to ask them for one, I need to give them two to three back, right? Like, Hey, I'm going to ask you for this piece of data, but I need to take two or three as part of that. And I think through following that strategy, it really did get buy. And as we went through it, I think overall I talked to, I had to do, I didn't have any true, uh, quantitative data. I only had qualitative data on this. I always screw those up. So if I did, whoops.

[00:12:21] Um, but we, we, I went and surveyed the engineering leaders and like, Hey, as we built the system work and the transformations we went through, what kind of gains usually in the general consensus between three to 5%. But when you're talking about 20,000 people, that's a lot of cycles. And again, it's already a fairly well run engineering organization. Making it better is extremely hard. And to get three to 5% across just through doing a migration, it's a better way of work. And we haven't even started leveraging AI with it. So when you're looking at that, that could be, that could be a double or triple number for a massive organization. Wow.

[00:12:51] And I was also interested in your idea of moving from a rigid map to a dynamic playbook through scenario planning. So how can organizations balance long-term street strategic direction with that agility that's needed to respond to this constant disruption we're seeing? I think the vision shouldn't change, right? I think that's going to be very calm for some, some companies, right? You know, we want to, you know, we want to be the world's largest networking provider. We want to embed security into the fabric of the network. That's not going to change. Like that's going to be a multi-shear strategy.

[00:13:19] But today with leveraging AI, like what I developed and I thought my strategy and I thought I had a good handle on it in August. August, it looks way different than it did in February and February even looks different than it did in March. And now it's, it's starting to get closer, right? I feel like we're, we're actually getting some ground on it, but like, I laugh when people come and tell me like, oh, we're doing our annual plan. Somebody approached me and said, we got to start our annual planning. I'm like, no, like the vision's still there. We still want to make, you know, obviously drive, you know, a better system work, you know,

[00:13:47] object for the team, more cohesiveness, engineering, more automation, all these things. Like the vision is still there, modernizing systems, et cetera, like how we do it. Now that's going to change, but let's plan a quarter at a time and iterate quickly. Right. I think anybody that, uh, and again, it's, it's when I go back to that statement around scenario planning, I know a lot of leaders, they go in and they, they want to double down to these things. And it's like, really? At this point, you don't know what the out, what the next three months is going to hold have three or four strategies. Hey, if this happens, then we'll go with this. If this happens, go to this in this happening.

[00:14:16] And if it goes to four or five, okay, that's a bit ridiculous, but generally two or three things. So that way, you know, you've got your vision correctly because you can pivot with the movements of the market industry, all the fun that goes along. And I think so much of those examples you just offered there will really resonate with people listening around the world. Are there any disruptions that you think many organizations still underestimate? Maybe. I think they're overestimating them to a certain degree. I mean, you know, like, uh, the last thing is the prime there.

[00:14:44] That's a SAS apocalypse that everybody talked about last quarter. And I'm like, guys, oh, you know, you're just going to do this all via cloud workflows. And I'm like, you still need that context. You still need these systems in place. And I think people are over-rotating thinking AI is going to fix every problem. And I think it's amazing. Trust me, don't get me wrong. I think it is the coolest thing I've seen in decades, but it's going to take a long time before it's literally just, you talk to a machine. It does all the, you know, the workflows and data and all these things. Like, you're still going to need that context, a human in the loop, right?

[00:15:13] Avoid these hallucinations. It was really what, uh, Mike Canterbrook said yesterday. It was like, yeah, you've got, you've got these different scenarios, right? And again, it's acceleration is context times intelligence. I thought that was a very, very profound statement because it is something we're going to do today is like, if you have that context and that intelligence and you can really move fast and it becomes about time to market, not cutting people, right? That's usually people think, oh, AI and you have to be able to know now we're actually getting products to our customers faster at a higher quality, better velocity, right?

[00:15:42] Through leveraging AI as a partner in how we build things, but it's still human involved. It's got to have the context. There are so much history. Cisco is a, oh my gosh, I always screw this up too. It's a 42 year old company. I believe. I mean, there's in the two main network operating systems. We, we are, I over, I hope oversee there's over four or 500 millions lines of code that are, that are 35 to 40 years old, right? AI has not figured out how I can go that large of a code space yet, right?

[00:16:12] It doesn't do an amazing job with it just yet. It's going to get there. It still requires a human to understand why we change this, you know, depth on writing protocols. Now, as we document more of this stuff, leverage more of a system more, create more of that context over time. I think that will become less of an issue, but it's years away. I don't think it's people are, you know, like, oh my gosh, everything's going to change. We're going to cut half our workforce. I'm like, it's good. But it's not that good yet. Yeah. And the thing with the SaaSpocalypse thing there is that organizations haven't got the

[00:16:39] time to maintain, manage, and update all these things because they're used to somebody else doing it for them. Exactly. Yeah. They say, you know, they go in and they're like, oh, well, you can do whatever. And I'm like, great. What's your roadmap for? Are you thinking of three or four steps? Because again, the folks here at Atlassian, we are about our products. Like, you have a multi-year roadmap. You understand the industry and the vision. Like, if you're doing it yourself, yeah, cool. You'll build a customized thing, but now you have to develop. You have to maintain it. I don't care if you're using AI or low-code development. Somebody has to put time in that.

[00:17:09] And it's somebody thinking about a roadmap, how you continually make it better. Right? And I think that's one of the things people, they get wrapped around. And I think at a small startup, I think, I will say that I think that's going to, that will disrupt some because it's Greenfield. Right? You can leverage AI. You can start from Roundup. And that will be a different way of working. We do need to lean in on that and figure out how that's going to happen because you want to understand the industry. But at the same time, I think thinking is all going to change the game for large enterprises or even midsize. I think it's maybe a couple.

[00:17:36] I had great conversations with folks at Google, Apple. They're not as far. Like, they're not, they haven't figured it all out yet, which is encouraging to me because I thought, oh my gosh, like Apple and Google, I'm sitting at a table with them. Like, so you guys have figured it all out. And they're like, no, you're actually, in some ways, you're slightly ahead of us. And I'm like, holy crud. Like, I am not alone in this problem. Great. You know, but I think it's, you know, it's, the opportunity is there. I don't think we've had the true value creation yet. I think we've seen some of it.

[00:18:05] Like my own development team, I asked him two weeks ago and I was like, hey, how are we doing with this? I haven't actually asked you guys because I haven't been worried about it. I know you're leaning in, but you got any metrics? Like, oh yeah, sure. Here, we've been tracking it. And I was like, okay, where are we at? And they go, well, we, we produced our resources by around 28% this year, but we saw a 65% productivity gain. Like, so you're, I'm seeing basically what nets out to a 2x gain just to, again, leveraging

[00:18:33] the AI code existence, leveraging the documentation and data that we put in to our context, right? It really is taking off. And now we've got goals of 100% AI developed applications in our tooling space because we do still have to do some customizations, forge apps, et cetera. But my goal for the team is guys, you know, put the context in the system, have it help you write the software, but don't, don't do the old way. Actually leverage AI to write the complete stack because over time we would need to gain experience in that. So yeah, in three or four years, we can go to bigger applications, right?

[00:19:02] So simple ones. Yeah, sure. That, that works great. You can refactor us, but as you get more complex, uh, you, you need to really have a better path there. And finally, there will be many people listening wanting to learn from the journey that you've been on. Cisco's transformation, anything you'd like to pass on to leaders that they need to understand before thinking about consolidating systems, introducing AI, or just attempting to create a similar unified systems of work. Anything you've learned, anything you'd pass on?

[00:19:31] First thing is be hardheaded, uh, and really, really double down on the strategy of, you know, make it a transformation on migration. I think if you, if you really go with that and understand how you want to work, not the way you have work, I think you'll be in such a better place. I think it's, I talk to leaders. I'm like, well, I want to work this way. Well, then do it right. Like this is your one shot. Really take advantage of that to transform leverage. It'll put you, it'll position you bigger for AI. Yes, you are. It's going to slow you down for a hot minute. It always does. I don't care who you are.

[00:19:58] Change management is hard in an organization, but after people latch onto it, the growth, and we've seen a complete hockey stick as people started to actually leverage things like Confluence, JIRA, automations, et cetera. We're seeing our usage just go through the roof, but it took some time to get there. It wasn't like overnight, like, Hey, everybody, Jason said, let's do it this way. Right. You're so you really do have to be consistent. State of your vision. Really think about transformation, not migration. I think that's a powerful moment to end on. I'll include a link in the show notes to your LinkedIn and all the work that you're doing at Cisco, but more than anything.

[00:20:27] Thank you for stopping by and talking to me today. My pleasure, Neil. Thank you. Wow. I think listening to my guest there is there's no pretending transformation is easy. There's no suggestion that AI is a magic wand that will fix every broken process overnight. And there certainly isn't an illusion that large organizations can simply flip a switch and suddenly become, Hey, we're AI native now. But the message behind today's conversation was very clear.

[00:20:53] If you want AI to deliver meaningful value, you first need clean workflows, shared context, standardized systems, and organizational discipline. And that point about transformation versus migration, that's something that will stay with me too. Because so many organizations fall into the trap of moving years of technical debt, insufficient processes, and fragmented ways of working directly into modern platforms without stopping

[00:21:22] to ask a much bigger question. And that is, if we were starting from scratch today, would we still build this way? Because there's a real opportunity here to reduce friction, accelerate decision-making, and allow skilled teams to spend more time solving some of those meaningful business problems instead of manually stitching systems and reports together. And as an old friend of mine, John Osborne, will use to say, it feels like it's built on twiglets.

[00:21:49] And honestly, hearing that engineer teams inside Cisco are already seeing huge productivity gains through AI-assisted development and still keeping humans deeply involved in the process, this all feels like a much realistic picture of where enterprise AI is heading. But as always, love to hear your thoughts. Is your organization treating AI as a genuine transformation opportunity? Or simply layering automation on top of existing complexity and technical debt?

[00:22:19] And if you had the chance to rebuild your systems of work from scratch tomorrow, what would you do differently? I want to hear all those stories. Good, bad, the ugly, whatever it is, tell me. TechTalksNetwork.com. Keep your messages coming over. A quick thank you to NordLayer for supporting the podcast and helping me make these daily conversations possible. And if you are listening and you're responsible for security or IT, you will know the reality.

[00:22:47] The reality of most of your risk now sits inside SaaS apps and browser activity. That gap is exactly what NordLayer is addressing with its new business browser. So instead of bolting security on from the outside, it builds it directly into the browser itself. This means you can control access, monitor activity, enforce policies and reduce shadow IT all from one single place.

[00:23:16] And most importantly, it does it without adding deployment headaches or complex onboarding. You get things like browser-based data loss prevention, SaaS access control and zero trust browsing, but delivered in a way that your team can actually use. So if you've been trying to simplify your stack while improving visibility, please check it out at nordlayer.com slash browser. Thanks for listening as always. Bye for now.