The Evolution From Network Engineer To AI Orchestrator w/John Capobianco

The ground is shifting under enterprise networks, and we’re not mourning the end of the engineer, we’re celebrating a bigger, bolder role. Automation has crossed a threshold: AI agents now watch the fabric, triage tickets, and surface fixes in plain English before a human logs in. We dive into that turning point with guest John Capobianco, AI and DevRel leader, longtime network architect, and co-author on Cisco’s pyATS, who shows how structured data and agent fleets turn toil into leverage without sacrificing safety.

We unpack the evolution from command-line craft to systems orchestration. John explains how parsing show outputs into JSON supercharges LLM reasoning, why read-only triage is the smartest first move, and how to design a supervisor agent with specialized workers for compliance, security, and config. We explore the very human side of change: fear of being automated away, the reality that automation usually increases your value, and the mindset to replace imposter syndrome with quick wins that your team can feel in minutes, not months.

If you’re wrestling with legacy equipment, risk tolerance, or executive pressure, this conversation brings pragmatic answers. We outline spec-driven development as a core competency for AI-augmented engineers, highlight the soft skills that now differentiate careers, and give leaders a playbook for responsible adoption: approved models, clear ownership, lab-first testing, and supported API access. Looking ahead, we sketch an agent-first future, natural language interfaces on devices, vendor-tuned models, and agent-to-agent operations that make networks safer and faster.

Ready to move from operator to orchestrator? Hit play, then tell us the first workflow you’ll offload to an agent. If this resonated, subscribe, share with a teammate, and drop a review to help more engineers lead the shift.

Follow John on LinkedIn: John Capobianco | LinkedIn

Referral Links

StreamYard: https://streamyard.com/pal/c/5142511674195968

ElevenLabs: https://try.elevenlabs.io/e1hfjs3izllp

Contact Digital Revolution

• Email: Jim@JimKunkle.com

Follow Digital Revolution On:

• YouTube @ www.YouTube.com/@Digital_Revolution
• Instagram @ https://www.instagram.com/digitalrevolutionwithjimkunkle/
• LinkedIn @ https://www.linkedin.com/groups/14354158/

If you found value from listening to this audio release, please add a rating and a review comment. Ratings and review comments on all podcasting platforms helps me improve the quality and value of the content coming from Digital Revolution.

I greatly appreciate your support and Viva la Revolution!

Transcript

Automation Crosses A Threshold

Jim 0:04

Over the past few years, something profound has been happening inside enterprise networks and infrastructure teams. Automation isn't just speeding up task anymore. It's reshaping the very nature of how digital systems operate. AI-driven engines are now monitoring, optimizing, and even resolving issues before human ever logs in. And that shift is accelerating faster in 2026 than at any point in the last decade. For generations, the identity of the network engineer was built around mastery of the command line. The command line interface or CLI wasn't just a tool, it was a badge of honor, a symbol of deep technical craftsmanship. But today, that model is giving way to something entirely different. We're moving from hands-on configuration to system level coordination, where engineers guide intelligent platforms rather than manually steering every device. And with that evolution comes a cultural tension that every technical team can feel. On one side, there's a very clear fear. If AI can automate the work that I've spent years perfecting, what happens to me? On the other side, there's a genuine excitement. If AI can take the repetitive task off my plate, what new possibilities does that unlock for me? This episode begins right at the heart of that tension. Because the truth is, we're not watching the end of the engineer. We're watching the role transition from something bigger, more strategic, and more impactful than ever before. Joining me today on the revolution is guest John Capianko. John, welcome.

John Capobianco 1:57

Jim, that was a really great. I got tingly, uh my ears are all standing up. That was an incredible introduction and what a way to um summarize this revolution we've been going through. And I love that the name of this show itself is the digital revolution. I think it's um important for us to recognize that this is going to be looked back at historically, I think, like other human revolutions, the agricultural revolution, the industrial revolution. Uh, we we this is a new agentic revolution that's happening right now, right? And I'm excited to be here. Um, I don't, you know, I understand that it is new, and you know, it's it's less than four years old, let's say, this this whole idea of generative AI, but the train has moved so fast in that short period of time. Um, I think it's important for us to have these discussions.

John’s Journey Into AI Automation

Jim 2:54

Perfect. And I appreciate that. Yeah, the revolution is is gaining speed, it's growing, and everyone's involved. And I think it's important to have uh guests like yourself on the revolution to talk about everything that we're uh seeing coming uh immediately and down the uh road here. Before we get into our conversation, could you do a favor and kind of provide a little bio bio about yourself, you know, who you are, what you do, and all that kind of good stuff?

John Capobianco 3:20

Sure. So my name is John Capo Bianco, and I live outside of Ottawa in Canada. Uh, I'm currently the head of AI and DevRail at Itential, um, who is a network automation platform and an orchestration platform who's embracing artificial intelligence in the form of flow agents and flow AI agents. Um, prior to my time at Itential, I was with Selector AI as a head of DevRail there for about 18 months. Cisco before them for about three years in an artificial intelligence role. And prior to Cisco, I was the senior network architect for the Parliament of Canada for about nine years. Uh, and that's really when the journey started for me, Jim, specifically network automation. So the we found that the greenfield environment that we had built at Parliament, building and assembling the network was one challenge, but operating it was a completely different challenge. And we realized that we were understaffed or at least very stretched in terms of staff. Uh, velocity was slow, and we needed to find a new approach. Around the same time, Ansible became popular in the network automation world and even some Python, different Python libraries. And that's when it really all started for me is when I started to do things programmatically in the network using infrastructure as code and YAML files and Python files. Um and when ChatGPT uh I applied for an API key to ChatGPT in December of 2022. So this is going, it'll be my fourth year of doing this come December. But I connected a uh a particular Python automation framework called Cisco Pi ETS. I connected that to the ChatGPT 3.5 API, specifically around testing interfaces. And it was like when it came back and said, yes, these three interfaces have problems, and here's why they have problems, and here's what you can do about those problems. Uh Jim, it felt like Gandalf when he touched the ring. Okay, I was sitting by the fire with my pipe kind of just muttering to myself, precious uh Pi ETS, artificial intelligence. It had a profound impact on me to see this artificial intelligence, and I can explain why. So to you know, not to boast or anything, but I am the co-author of the Pi ETS Cisco Press book. Um, and I'm I'm one of the experts who use this uh automation platform. And prior to testing interfaces, the old way used to be writing a test per use case. And to test interfaces on a say a Cisco device, it was like 19 to 30 different discrete, handwritten, brittle, fragile, bespoke tests given a certain platform and given certain variables. You know, it was hundreds of lines of Python. Now it was doable and it was very valuable once you had written that code. Um, but for me instead, just to take the payload from the show interface command and send it to the AI and say, you know, tell me if everything's healthy without writing tests, without writing any of that code, and for it to come back with a high quality answer, uh, really changed my thinking. It changed my thinking all those, you know, years and months ago. It it really had a profound impact. Uh so that's sort of where my my origin story is with artificial intelligence.

Life At The CLI And Its Limits

Jim 6:57

You know, that's a great point. You know, we really think about with network engineers and we look back, you had mentioned a four-year period. You know, you go back prior to four years, maybe you go back maybe you know, five to let's say five years to a decade. You know, a network engineer, their role is much different. Can you talk a little bit about you know what that network engineer was, let's say in your in your early part of your career? What did what were you doing back then as a network engineer before AI?

John Capobianco 7:24

Yeah, so there was a lot of um everything was done at the command line. So you had to have uh a broad knowledge of different show commands, configuration commands, modes of operation in a device, one device's version of the commands compared to another device's version of the commands. They weren't uniform, multi-vendor. Um, you know, logging in and configuring things from scratch. Usually how we used to do it was like a notepad file per device. And in that notepad file, you would have your procedures and your configuration commands and everything all drafted up per device per file. And some operator would log into the device and apply the config. And then testing was a whole separate round of things that was done after the fact. Everything took a lot of time, everything had friction. You had very small contained change windows to achieve big goals. Sometimes it might be swim software image management, you know, just upgrading a router's iOS device or whatever. Sometimes it's configuration changes, sometimes it's reacting to an external event. Some component fails, some wire gets cut, some power gets lost, right? There's external factors around the network that sometimes you have to react to. Um, but artificial intelligence has really accelerated all of those things. And I really do try to build myself like digital coworkers or digital uh subordinates in a way, where my expertise of 25 years of doing network automation can be poured into as skills, as model context protocol tools, as retrieval augmented generation. There's a variety of ways me as a human can translate that knowledge into a digital coworker that then I can just ask to do the thing or the status of the thing or the health of the thing in natural language, right? It's just another layer of abstraction on top of other layers of abstraction that we've gone through as humans, right?

The Rise Of The AI Orchestrator

Jim 9:44

Yeah. Yeah, John, I appreciate you uh sharing, you know, that those genesis points, you know, where you've seen the and experience of the automation along with you know software-defined networks and artificial intelligence really becoming that reshape uh your responsibilities. And I love the analogy regarding Gandalf, the first time he actually physically uh touched the one ring. I think that's a great way to kind of uh uh frame everything when it comes to the fundamental change and how everything was evolving because of AI. Now, let's talk about the new role. You know, we have the aspect of a network engineer, and and and you're uh what you had presented to me when we were talking prior to the podcast episode here was that that network engineer is more of an AI orchestrator, and that role was kind of evolved in that aspect of it. And I want to talk a little bit about, you know, what that AI orchestructure, I'm sorry, orchestrator actually does. You know, you're moving away, and you think about it, we're moving away from that device level troubleshooting. Now it's into managing intelligent uh systems that are and they self-optimize and everything like that. Um, and also I think looking at how AI also how it overlooks and watches, you know, workflows, you're gonna have anything from you know policy engines to autonomous network functions and things like that. Can you talk a little bit about that new role of the AI orchestrator, you know, what they do?

Agents At Work And Culture Shifts

John Capobianco 11:13

Well, I think it is a human resources problem as much as it is a technology problem. Um, meaning if if I were going to build a team of people under me right now, and let's just say I had unlimited budget and I could build experts to do certain things, how would I organize them so that I, as the human, could just kind of be on the loop or in the lead of these agents where I'm still in charge of them as a fleet, as a workforce, as a as a force multiplier. But they are autonomous, but they do still have enough free reign to be valuable to me. Um, some might do configuration, some might do compliance, some might do security, each of them with their own set of tools and their own set of skills. And you maybe have a supervisor agent that you report to, the one interface with a human, and that supervisor agent has a fleet of other agents under it. Um and I and I sort of I'm looking at the current present state of agents right now. Six weeks ago, or before Open Claw, I'm not sure if you've been following the Open Claw craze, but Clawed, and he got sued, so he changed the name, and now it's become one of the number one GitHub repositories on the internet today, and there's 1.8 million open claw agents on Earth. Um, that should open people's eyes to this new idea of an agentic era and the digital revolution that we're living in. I think we are the last generation. 2026 is probably the last generation of purely biological beings in the workforce, particularly in IT. I think Gartner pegs it at 70% of IT-related activities will be AI augmented by 2030. So that's only four years away. They believe 70% of all things done in IT will have a component of artificial intelligence being involved in that activity. And um that might even be um uh a timid number, right? That might be 2028, right? Who knows, right? But the way things are moving, um and and and I still think it's very democratic, right? That 1.5 million agents, uh, that's just curious people, right? They just followed the README and set up their own AI agent. It wasn't very hard for them. People of all walks of life have agents running now. It's not restricted to people who are network engineers, right?

Jim 13:53

Yeah, and right now, you know, we're really faced with uh a huge cultural shift that that's happening with inside technical teams. I'm gonna ask, you know, your experience and also too, you know, what your um ideas and also two uh thoughts are related to when it comes to technical teams, um, we're seeing the reaction of, you know, when AI begins taking over all the routine tasks, you say, like with the IT function and stuff like that. You know, how are teams currently in business organization, even into government side, you know, how are they reacting to AI taking over these routine tasks that they have?

From Fear To Value In Automation

John Capobianco 14:31

Well, I I think it rhymes with what we did in network automation, meaning there was a big fear and an immediate backlash to network automation that you're going to automate yourself out of a job. That that that sort of has hovered over network automation for a very long time. This idea that you could automate yourself out of a job. Personally speaking, every time I automated something, even big things, there was no impact that they didn't suddenly let go of me, right? They didn't say, well, thanks for writing this code, and and you can, you know, we don't need you anymore. I was actually always rewarded for automating. What else can you automate? Where are the other opportunities? Where else do you think this could go? And it sort of became John was sort of the one in charge of that code base, and he was valuable to the organization now because he was one of the few that understood the new network automation code that we were doing. Not to say I hoarded that knowledge. I had training sessions and other people were aware, and those people became more valuable as well. I think if it's John's agent that people are talking to, John is still a very valuable person if he's the one building and maintaining and augmenting and responding to demand to improve that agent that he's built for the enterprise. Right? You might have an agent that you put in Slack that people can just talk to. How's the health of the Wi-Fi in Atlanta? And that agent just answers them with the data that John builds into the capabilities of the agent. You made the good point that it's a shift, right? That they're now digital architects around agents. And can I, as a human with the experience and the certifications and the stuff I've done in networking, make a super quality agent so that people don't have to bother me as a human? It's not about bothering me, but it's about could I publish that in our Teams directory or you know, whatever platform it is that we all communicate with? Could I give it an email address? Could it tap into my Service Now tickets where I have a category in ServiceNow that says agent? And if that category is selected, that ticket goes to the agent first for triage. That's how we're going to augment things, Jim.

Jim 16:52

Yeah, it's uh is as I mentioned earlier, you know, it's a cultural shift, but really it's a mindset shift from protecting my job to elevating my impacts. And you've definitely uh capitalized on it and dialed into that. Now, again, I I do recognize the importance of psychological safety, you know, when it comes to one's career and job. Um, and I do know that, you know, AI being a disruptor, um, and it is assisting in a lot in operations. But I think when it comes to leadership, both on the IT side and also the corporate organizational side and even in government itself, you know, how can what are some of the recommendations you would have for leaders to kind of help reduce that fear and really kind of build some confidence in this new model that we're seeing evolve in front of our eyes?

Stability First, Then Safe AI Triage

John Capobianco 17:45

So I think that um it's everyone's so it's so early, Jim. It's so early that it's still an opportunity for everyone, everyone listening to this, let's say, to emerge as an AI superhero in their organization. Right? If if you bring the hardest questions with some answers, right? Uh why aren't we using AI yet? Because we can do it on-prem securely with open source models, or we have an agreement with Anthropic or with Azure or with Google or with whoever, right? So, how can I make an impact on my team to be the champion of AI and learn how to use the tools and learn how to solve some problems with it and have meetings about it and share that knowledge and share it with your leaders? Leaders right now, the best thing they can do is empower people because I don't you hate to see the progression go from it's not allowed at all, AI, to highly restricted use cases. You can use it to actually, it's generally available for employees to use. To it's highly encouraged that you use this tool all the way to you must use AI. That's sort of the spectrum that we're sort of racing through as enterprises. And now you have to align in your mind where your enterprise is at, because eventually it's going to become an expectation that you're using the AI that the company's invested in, or that the government bodies invested in, or the tools that they're spending the money on tokens. They want to augment the workforce, they want to increase your productivity, make your job easier, give you access to better sources of data. So you're so everyone's job is going to change. And I think that's sort of the the that spectrum is of no AI allowed at all, to you must be using AI, and somewhere in between is what we face right now as humans.

Jim 19:47

Yeah, I do agree completely with that. Let's talk about the practical challenges that engineers today face uh during this transition. And you know, when I we think about the accelerated evolution that's going on right now related to AI, is network engineers, or however you identify in that IT sphere, it is very difficult as humans to let go of what used to be, those old habits, you know, writing code and everything like that. And also, too, I think when it comes to really accepting that you need to trust in AI, especially related to things that you're going to be doing with it for it to give you recommendations or even automation and things like that. Um, you know, what are some of the practical challenges above and beyond those two that you think that engineers are facing in this transition?

Bridging Legacy With Modern AI

John Capobianco 20:42

So networks have to be stable. Stability is always going to take precedence over innovation on networks because of their criticality, because things like 911 ride on them, because security cameras and phones are on them, because that's the way everything in the world is accessible is through their IP address. So I don't mean to be hard on network engineers that maybe haven't done automation yet or haven't started AI yet. I I'm not judging them. I understand the situation they're in and that they don't have the luxury to innovate. It's more of a science than an art. Whereas the world of computer programming and application development, you have a little bit more leeway and flexibility there. It's more about design decisions. With a network and with infrastructure, right, everything rides on that network. So so what if we're not using AI, right? Is the are the packets flowing? But I like everything else, there are huge advantages, and there's huge. I just did a demo right now. I just got off of a call with a demo where you open up a ServiceNow ticket. I sort of alluded to this, and you mentioned that an AI should triage it. It triggers an agent that connects. To the device and does the thing and runs the tests and comes back with a report in Service Now, updating it live, right? With enriched text and full test results and recommendations on how to solve the problem within minutes of the ticket being opened and without a human being bothered whatsoever. Right now, I think that is remarkable. And it's read-only and it's safe and it's human on the loop. It's all the wonderful things we'd ever want. Imagine starting with triage with tickets that come in, and those tickets just get triaged by AI. Yes, most of them are still going to filter up to a human to resolve them. Maybe that's your next step is classifying what tickets resolution could be handled by an agent safely or with minimal risk. Are there certain things like shutting a port or adding a route or doing whatever that would remediate the problem? Then you have the AI do. Right? So we as humans are going to have to start moving into the business of building these agents to do this plumbing, right? I don't think we as humans need to worry about memorizing show commands and memorizing the nuances of the CLI. Um, you know, there's certain things I think we can abstract to human language.

New Core Skills And Spec-Driven Build

Jim 23:23

Yeah, and I think also, too, you kind of keyed on another challenge, you know, the aspect of when it comes to the tools and the resources out there, if you're not really educating, informing, and enlightening yourself on what's available out there, you know, these platforms are evolving, you know, rapidly. You know, I go back and look at the beginning of you know, the digital revolution as a podcast is in the third season to see back when I first started those first episodes of what was available out there related to AI and see just in two and a half years later, roughly, where things are today, it's it's less like light years difference. One other thing I wanted to ask you, you know, related to on the network side, and this is something that people ask, I occasionally I'll get this question where you have a lot of companies that still have legacy systems. And it's kind of like in the Dilbert thing, where they want, you know, the executives want to keep the old analog system and connect the digital system to the analog system and have one big massive data system. You know, when it comes to working with executives and things like that who really want to have some aspect of legacy with these next generation AI things, you know, how does how do you recommend coordinating that um with the network people in working with the executives, maybe at their companies or their organizations?

Roadmaps, Curiosity, And Quick Wins

John Capobianco 24:46

Well, I think there's there's AI for networks, right? Which is sort of what what I'm in the realm of using these tools to interface with networks. And then there's networks for AI, which is sort of a different discussion, building the backbones and building the hyperscaled GPU farms and the back-end networks and fabrics and some of the new protocols. So there's a lot to consider as an executive uh whether or not you need to start investing in GPUs and racks of new equipment and the network expertise required to hyperconnect all of that, or if you're just going to consume uh you know and do inference where you maybe don't need those GPUs, right? Um the other thing is in terms of legacy networks, I'm hoping we're approaching the era where at least all of the network equipment out there has REST comp capability where you can use REST APIs and Yang models to interface with the network equipment as a standard. Um, most network equipment also supports things like Pi ETS parsers. Uh it is multi-vendor and it will work across a wide variety of equipment to turn the show commands into structured JavaScript object notation. And that's a big key. And it's been a big key for my success with AI is not just sending raw text and raw command line output copied and pasted, but parsing it into structured JSON because as far as the LLM is concerned, it can read JSON universally. It doesn't make any difference that it happens to be network key value pairs, right? It just it's just structured data that it understands. So my my quality was really good because I was using and still use that structured JSON to improve the inference, right? So as long as your platforms can support either parsing of some kind or a rest interface, um, I don't know, you know, now legacy a system that's 20 years old, it you know, it should still have some of those capabilities. I I don't think that we should dismiss uh you know previous generations of hardware from being able to use AI to manage them.

Jim 27:18

And we've talked a lot about uh the challenges and you know where things are kind of headed for you know AI augmented engineers. Um now I think we definitely need to talk about the skill set. So when we're looking at core competencies, some of the things I threw out were you know automation, you know, scripting, data literacy, um, model awareness, and also systems design. Are there other core competencies that I just didn't mention that are important to an AI augmented engineer?

Leadership, Guardrails, And Ownership

John Capobianco 27:51

So I think that less and less we're going to be worried about Python and and or Go or Rust or C or programming languages. I think that something that most should start paying attention to is something known as spec-driven development, SDD. And there's a GitHub repository to get you started. There's lots of information on the internet about this. And it harkens back to TDD, test-driven development, but it's spec-driven development. And the spec files are just natural language markdown files. There's a structure to this, though. And that's how I think systems are going to be built moving forward. So network engineers need to start to become very proficient at writing specifications that an LLM is going to turn into applications or into networks in my case or into infrastructure. And it's perfectly aligned with network infrastructure because network infrastructure has always been spec-driven development. We've always used specification files of a kind when we design and architect these networks. So we can take those architecture specifications and turn them into something that generates the configs using the LLM and this spec-driven development approach. So I you might be hearing this first for the first time here. I think that's going to become a critical skill to the AI augmented engineer of the future, is being able to interface with an LLM through specification files.

Jim 29:31

Oh, very good. You know, looking at those core competencies, obviously those are very critical. But an AI augmented engineer is going to become more visible, more front and center. So what I highly recommend, uh engineers that are looking at the aspect of AI, is you got to work on your soft skills. You got to be a good communicator. You have to be able to do cross-functional collaboration. And you also have to work on the strategic thinking aspect of yourself because you're going to be called on to do a lot, very different and unique things, expanding and going above and beyond what a normal legacy engineer would when it comes to network. I also think that you have to have a level of curiosity and adaptability. Um, as you mentioned earlier, um, that that aspect of curiosity and adaptability is going to matter more because it's not about memorizing commands anymore, writing code, memorizing uh commands and things like that. Question I have for you related to here we are in 2026. How can engineers build a personal learning roadmap for their future? What would you recommend as a roadmap?

John Capobianco 30:40

Well, I would augment that. Honestly, I know it sounds funny in a way, but I would use AI to augment that story. And I would in particular look at either Chat GPT 5.2, uh Gemini 3.1, or Claude 4.6, any of those at all. And honestly, be honest with yourself, be honest with it, lay out your goals for year one, two, three, four, five, right? What you want to achieve, where you want to go, where you are at right now, what what skills you don't have, and work with the AI to develop that plan. Um, now, five years is a very long time, right? I would even say a one-year plan where in a year, could you develop an application using AI? Right? Could you actually build something? And a year even is long. In this rapid era, it's probably days away from you starting your journey and actually having a hello world application. And and try to find something that will improve your life. Could I make an AI agent that can check the health of my network device? Send me a Slack message, send me an email, uh, run tests, generate configs, whatever it is that you do on a day-to-day basis, right? Just think of the top five things that maybe you could offload to an agent. Are there things that are there good candidates for things that you could maybe truly yourself write the agent to offload that work using your own domain-specific knowledge? I would say yes. I would say everyone listening to this probably has five things a week that they could offload to an agent uh with a high degree of success.

Mindset Shifts Over Imposter Syndrome

Jim 32:33

So we've talked about the importance and obviously the skill set and also to um the important elements related to AI, an AI-augmented engineer. Now I think it's important to kind of move towards the leadership that obviously is going to be interfaced with these AI augmented engineers because they're going to be the enablers of the transformation. Um, one of the things that a lot of uh uh of uh uh of uh consultants, but also to subject matter experts related to AI augmentation, uh, they're pushing really hard out there to create that culture, the mindset that businesses today need to really look at this AI enabling um not just as like, you know, running kind of a we're gonna run a pilot, to look at this as an opportunity for them to really grow the enterprise and to take off a lot of those safeguards. And what I mean by that is that you should want to have failure happen because you're gonna have a better outcome. If you have sometimes missteps that lead to a better outcome, it's gonna be better for the company. Another area, they need to really invest more in the training, even mentorship. You know, John's like someone like yourself is an incredible mentor, but have that structure of upskilling paths, not only for AI augmented engineers or engineers that are in training to become that, but also others that will work in and around and collaborate with them to really have that understanding of AI. You had said earlier, John, about HR, about, you know, related to more of the job descriptions and the career side of it and everything like that. There has to be a career path that's established for AI augmented roles. What let me ask you this, you know, how can leaders, you know, what would you recommend for a model related to their behavior when it comes to you know having their teams understand that we want you to adopt and to implement AI? What would you recommend as a model for those leaders?

Practical Steps To Stay Relevant

John Capobianco 34:44

So I think I think business leaders, the number one thing that they need to define is this idea of responsible, responsibility. Who's responsible ultimately if the AI gets it wrong? Is it the person who wrote the code? Is it the person who wrote the agent? Is it the team behind the project? Is it the LLM who hallucinated? And we blame the partner or the model provider. That is something we need to sort out so that way individual developers feel that the company supports them in exploring these solutions without maybe the finger pointing associated with, well, John's agent did X, Y, and Z in the database or something, right? Clear path for success. Again, it rhymes with network automation. Do this offline, do this in a virtual environment, do it in a physical lab, do it anywhere but production first, and you learn your lessons in those lab type environments, right? Don't do shadow AI. Bring your concerns to the company, right? We need an approved model. We need API keys issued by the company, paid for by the company, supported by the company for us to develop with, right? Don't spend the $20 a month and use your personal Chat GPT API key to get some of this done as a shortcut, right? You actually have to have those difficult conversations with your leadership. And leadership, you need to be thinking about this in the reverse way. Have you made AI accessible to your teams? Are there guardrails? Are there governance? Are there training? Do you go through the half-day exercise of here's the model that's available to you, here's how to get an API key responsibly, here's what's in, here's what's out, here's what documents are approved for RAG, here's what documents are sensitive and not approved. All of that. There's lots of opportunity here for leadership to do exactly that, to be leaders in how their companies move forward with AI. Uh, and I think all of those things will lead to higher adoption rates, which leads to better code, which leads to faster delivery with more velocity and more revenue, right? It's all tied together.

Jim 37:01

John, I think it's real important at this point of the episode, you know, as we're kind of winding this episode down, to get your advice, you know, for the network engineer, the engineers out there that are looking to get involved related to AI, obviously, and really for technical professionals. Let me ask you this question. You know, what what's one thing that you wish you had known earlier in your career as to where things are today related to AI?

John Capobianco 37:28

I I really wish I didn't have so much imposter syndrome when I got started with AI. Even to this day, um, I've I've been very hesitant to call myself an expert in AI. Uh now Google developer expert. I am a Google developer expert, so I guess I kind of am, but even now I'm a little hesitant to say that because I my background was in computer science. Yes, I did computer programming, but I didn't do computer science. I don't have any idea what these algorithms are doing or what they mean. I don't have the mathematical background that I thought was required to break through to AI. I thought there was this mythical glass ceiling holding me back from actually, you know, being a consumer of AI and using the tools to help my career with network automation in particular. But I thought there was this kind of mythical ceiling that you really needed to have a certain background or a certain skill set or come from a certain educational background to start using it meaningfully. And in fact, if you can do basic REST API stuff, it's accessible to you. And it's more accessible now than ever with things like MCP and things like AI agents and things like Claude Desktop and Cursor and uh Gemini CLI. So don't let that hold you back. If that's what's holding you back, this idea that it's somehow more difficult than I would ever think I could tackle or figure out as a human, or you know, do you know what I mean, Jim? I don't know that I'm articulating this well.

Five-Year Vision Of Agentic Networks

Jim 39:08

You definitely are, because I was going to ask, you know, what was a mindset shift that can help accelerate the transition? And I think to your point of saying that imposter syndrome, because even in my own experience related to artificial intelligence and working with guests and also research and talking and you know, doing polls and surveys and everything like that, I find a lot of that the aspect of the imposter syndrome in there that, you know, I can't do this AI stuff because this, that, and the other thing. And I think it's the aspect of the head trash. You've got to eliminate that because it's a whole new realm of possibilities. And until you fully try and understand exactly and you immerse yourself into this new, these new changes, you don't unknown what you don't know. And I think that's important to your point as an imposter syndrome. What are some of the practical steps that engineers, uh network engineers can take right now in 2026 to stay relevant and thrive in this evolution?

John Capobianco 40:08

Well, stay curious and be inspired and not like I think of the meme of the person looking out of one side of the bus and it's happy sunshine, and the person looking out the same same bus but out the other window and it's a rocky, dark, and gloomy. You have that decision to make every day. And every time you hear about an artificial intelligence, to look at it as an opportunity to learn something new, or to continue to dismiss it. Like I can't change the hearts and minds around that. I I hope my message has inspired you to be more curious and to take your first steps with these tools. But honestly, try to find that thing that could help you. I think it's an on and off switch. Once you actually see it do something of value, then it becomes valuable. Right now, artificial intelligence isn't valuable to a lot of network engineers because they haven't tried to do anything valuable with it. They they they chat and chat GPT for 10 or 15 minutes and they sort of lose interest or don't see the hype or don't see what the big deal is. Try to do something meaningful with it, which may mean you need to write a little bit of code to get started. But uh the you're gonna be rewarded hugely. You know, they're gonna get huge rewards from jumping onto this train now because it's moving very fast. And the longer you wait, the further away that train's gonna be when you do decide, you know, and maybe it's not your decision to make. Maybe someone's gonna make the decision for you that you do have to learn AI, and that's you do not want to be put in that position either, right?

Jim 41:46

Yeah, you're 100% on that. Because, like you said, you know, we've seen, we were there at the you know, you don't have to go far back when that train started to get speed, build speed. It's gonna get further away, it's gonna be moving really fast, like a bullet train, right? The other thing, too, if you're gonna wait until a company, your leader of your of your you know, team or maybe the company is saying, hey, we want you to upscale, we want you to move into this role, you're gonna be not in a good position, to your point. You got to take the initiative and embrace it and move forward with it. Because really, John, I mean, this whole episode, what you've laid out here is that the future really belongs to those who can orchestrate, not just operate. Operate is the past, orchestrate is the future. So I really appreciate you laying everything out on you know, on this role changes that are happening, the evolution, and also too, like you said, the speed of that train when it comes to AI augmentation for engineers. You know, to close out this episode, you know, I really view that there needs to be that human AI partnership. It needs to really, I think it has started, but I just want to just say formally that it must happen. It must start now. Because when you think about an engineer's role isn't disappearing, they're not fading away, they're revolving and they're evolving upward. It's it's it's a rapid, accelerated evolution, and evolution happens no matter what. You either grow or you don't grow and you fade away. And also, AI is a force multiplier for human expertise. And what I mean by that is it's not, it's going to augment you. It's not going to replace you. Yes, it's going to replace those uh repetitive tasks and things like that. You said it earlier. It's those who can really take that AI and do creative things with it and make it do unique things and things that are important to a business or to an organization or a government. That's super important. And I guess uh my last question for you is you know, uh, we're talking about a vision that's you know high performance with these AI organizations. Orchestrated teams. What do you think it's going to look like in the next, let's say, five years?

John Capobianco 44:08

Well, uh five years, I think that the teams are I don't think they're going to be smaller. I think they're going to be very different, though. I think the way we approach solving problems is going to be agentic first very soon. And there's going to be sort of a holding pattern on, let's say, uh expanding human teams, right? So in five years, 70, 80% of infrastructure is going to be augmented by artificial intelligence. Uh, I think we're going to have uh an exponentially larger number of data centers around the world and neo clouds and different types of cloud providers. I think there's going to be more specialized models. I think we might see more custom models that are maybe vendor supported. You know, maybe the vendors of the world put out their own language models to give us an even better way to interface with systems through natural language. Uh, I think MCP is going to lead to a lot in the next year. I think A-A, agent-to-agent communication, is going to lead to a lot more. I think that the whole idea of a gentic infrastructure is going to dwarf the World Wide Web. I think it might actually be the first thing to embrace IPv6, where agents are given their own IP address or agents can participate in the routing protocols, where your neighbor might be an AI agent. And by neighbor, I mean your BGP neighbor or your OSPF neighbor, where the agents can actually peer with routers and have even better insights and make even better decisions on routing. So I don't dismiss anything. I think that on right on the hardware that you get of the future, the servers of the future, the routers of the future, they will likely have GPUs or TPUs or DPUs, small, little purpose-driven pieces of hardware that let those routers themselves have a natural language interface. Right. So I see the pro the proliferation of natural language across, you know, and and maybe we we started on the CLI, right? And the CLI is dead, long live the CLI. You know, maybe we actually will break away from the shackles of the CLI in the next five years and move into a purely natural language-driven uh infrastructure.

Where To Follow John & Closing

Jim 46:44

Very good. John, how can people follow you, communicate, connect with you?

John Capobianco 46:50

Yeah, sure. So um I've started something called the Vibox Forum. And this is all we do is share AI ideas and share code. We don't call each other's work slop. It's a safe, inclusive space that you can reach out, and I'm I'm always in there. Uh LinkedIn is a good place to find me. Twitter X is a good place to find me. My YouTube channel's always got something interesting on there. And I love and cherish connecting with people from all around the world. Uh and and so feel free to reach out to me.

Jim 47:19

Beautiful, beautiful. And uh, if you could do me a favor after we're done uh with this session, uh if you could send me the links that you want me to provide, and then I'll put that into uh not only onto the Digital Revolution uh YouTube channel, but also obviously in the audio podcast uh description uh for this episode.

John Capobianco 47:39

All right.

Jim 47:41

All right, John. Thank you so much. I greatly appreciate you being on the revolution. And everyone, please continue to stay uh curious, informed, and also always continue to check in with the digital revolution. Thank you, everyone.