The AI Mindset Shift Every Network Engineer Needs

Dakota Snow• 00:00
John, we are here at Cisco Live 2026. I am so excited. Like the vibe in the air right now is just electric. You know, what you guys have been doing here at the ViveOps Labs, I just caught one of your sessions, and it’s super cool how you guys are really demonstrating, I feel, like, the future of networking and what it’s going to look like.
John Capobianco • 00:20
Well, thank you, and I appreciate you joining the session. And I agree with you. We’ve got the shirts to match the vibes. We’ve got the ViveOps lounge here live at Cisco Live for the 1st time. And it’s an open-air, free drop-by. We’ve got sessions going every other hour. We just finished up a session about model context protocol.
John Capobianco • 00:37
And you know what? Someone came up to me after and said, this was so perfect because we are investigating how to get into MCP. So it was really timely content. And everything around Cisco Live is AI.
Dakota Snow• 00:49
Yeah.
John Capobianco • 00:50
This may need to be called Cisco AI Live next time around.
Dakota Snow• 00:53
You know what? Last year, I feel like was the 1st year where Cisco really pushed heavy into AI and everyone was just throwing AI on everything, but there was no rhyme or reason behind it. Like no one knew what truly what AI was going to do. But this year I feel like everyone’s prepared. They’re like coming with demonstrations like what you guys are doing and actually showing, hey, we’re not just talking about AI, we’re showing how to use it, which is a super powerful tool.
John Capobianco • 01:20
Well, I agree with you, and I think that I don’t want to say overnight, but the world has dramatically changed in the past few weeks or months. We see an explosion of model context protocol from vendors, from open source, from individual contributors. We just did a blog post, and there’s already, I would say, 60 plus viable infrastructure model context protocol servers. 60 of them. That’s crazy. And the protocol is less than two years old. So, if all of this sounds like sci-fi to you, just take a step back, take a deep breath.
John Capobianco • 01:55
We are here to help, and we’re gonna go slow. Because everyone needs to go along this journey. We don’t want to leave people behind, right? So that’s part of my message. It’s part of what Atential is trying to do. It’s the whole spirit of the Vibe Ops Lounge is to try to give people practical steps on how to do this in production. That’s amazing.
Dakota Snow• 02:15
For network engineers, what do you think is the easiest way to break into the world of AI and using MCPs?
John Capobianco • 02:24
So I actually really think that finding one that applies to your day-to-day life. Let’s say a source of truth. We’re looking at InfraHub, who is right across from us here in the Vibops lounge. They’ve got an MCP server. So why not say, here’s my IP addresses, my IPAM, my old Excel spreadsheet. Please populate it into InfraHub. Via MCP with natural language.
John Capobianco • 02:52
That’s not hard to do. It really isn’t. That’s about a 20-minute investment of your time.
Dakota Snow• 02:57
And back in the old days, when we were going to do something like that, that’s like a week-long project potentially. And hundreds of lines of code.
John Capobianco • 03:05
And it’s all brittle, and it’s bespoke, and it’s custom, and it’s a lot of it’s throwaway. You’re only going to do that once, right? Whereas the MCP, you plug it in like a USB key, and now you can literally talk to those external systems.
Dakota Snow• 03:19
You know, it’s truly the future, and I feel like network engineers who are adopting it now, like I still feel like we’re still in the early phases, we’re still in the discovery phase of what we can do with it. That’s really gonna unlock the superpower for network engineers. You know, we’ve had so many iterations in networking and different technologies emerge, and it’s scary at 1st , but as people start to realize and use it, it’s just unlocking like this heightened level of expertise where it’s now accessible for us all to do.
John Capobianco • 03:54
Exactly. And it’s not just about speed. I mean, we all want to go faster. But the network is still the bottleneck because the network touches everything and we need to be very careful with change management controls, change requirements, approvals, all that stuff. I think that now with AI and MCP and agents, infrastructure can have a seat at the table of the business requirements and solving business problems and being able to deliver faster with higher quality, with less mistakes, and to remediate faster when there is an outage. The outage cost per minute is now like $12,000 is the average minute downtime, something like that. But what if it takes you an hour to solve the problem?
John Capobianco • 04:41
Right? Right. So there is real value in adopting and embracing the new way of doing things.
Dakota Snow• 04:47
Now, what about, I know there’s still companies that are definitely afraid of AI. What do you say to those companies and those organizations that aren’t willing to adopt it?
John Capobianco • 04:57
Well, I think it needs to start with leadership, but I think that being aware of your options. So a lot of people are skeptical or a little concerned about sharing their corporate data, their intellectual property, with a cloud model. Because what if they train on that data? What if they learn from that data? What if they can give one of your competitors access to your data? So that has left a lot of people, they want AI. It’s not that they don’t want it, but they want to do it responsibly.
John Capobianco • 05:27
And can we do it locally and privately? Well, open source models have come a long way. Now there’s still a bit of a parity gap between the foundational models from the hyperscalers and open source models, but that gap is closing. So what I would tell those companies is empower your staff with open source, free, private, local models. Do it privately, do it air gap, do it controlled, and start with read-only exercises. If you were to hire a new junior engineer, what’s the 1st thing that you tell them to do? To go learn, to go investigate the network.
John Capobianco • 06:03
Right, right. Go learn and probably update documentation. Exactly. Right? I know it’s a crappy job, and every junior’s ever had to do it. Hey, welcome to the company. Here’s the SSH key.
John Capobianco • 06:13
Go update all the documentation. Well, what better job for an agent? And it’s similar to a junior engineer. I’m not going to give my AI agent full control to change BGP policies on day one. I’m going to start with documentation, testing, compliance, maybe triage, ticket enrichment. There’s so much you can do that is safe, that doesn’t have to change anything. You know, it’s like people for 10 years now have conflated network automation with configuration management.
John Capobianco • 06:42
Right. They’re not the same thing. Yes, you can automate configuration management, but it’s only one small aspect of automation.
Dakota Snow• 06:50
Well, and there’s this misunderstanding. Everyone thinks you got to like, if you’re going to adopt AI, you got to do it all. Like, you got to give it the full keys of the kingdom and do everything. You don’t have to start there. You can start small. And speaking of, you know, you mentioned that companies that are wanting to adopt AI, they should start with a small local model and things like that. Can we even break that down more for people that really want to try to go pitch this to the leadership team?
Dakota Snow• 07:15
Like, how can they convince the bosses, like, let’s start smallest, different codes into the water?
John Capobianco • 07:20
Right, so I would, there’s three open source frameworks that I would look at. The most popular being O Llama. Not Llama the model, O Llama the framework. You pull it down locally onto your Mac. I would recommend the Gemma 4 model from Google or the Phi 4 model from Microsoft, Mistral, DeepSeq. There’s many options to pull the model onto your local machine and then start to incorporate it into your day-to-day life. Now that local model has a REST API endpoint.
John Capobianco • 07:51
So you can incorporate it into things like Ansible or Python scripts. So the on-ramp, right, I download a llama, I download a model, I turn on the REST API, and I write a little Python script to go get interfaces. And tell me if they’re healthy. That’s a perfect day one hello world exercise that you can try at home.
Dakota Snow• 08:11
Now, you mentioned recently that you guys did a blog, a blog article, talking about all these different MCPs out there. What excites you the most when it comes to MCPs? What are the cutting-edge use cases you’re finding?
John Capobianco • 08:27
Well, I think because it’s a protocol, I think I came up on protocols: RIP, OSPF, BGP, HTTP, SMTP, NTP. Now, where would the world be without SMTP? Yeah. Where would you be? Where would you do without HTTP? Absolutely.
John Capobianco • 08:45
So the fact that it’s a protocol, and most network engineers, if you say it backwards, it’ll make more sense. It’s a protocol that provides context to a large language model. That’s all its purpose is. It’s JSON RPC2 under the hood. So if you’re real nerdy and you need to know that, under the hood, it’s JSON 2.0 RPC. And it’s client-server. So the server is going to advertise: if you were the client and I’m the MCP server, I’m going to say, hey, Dakota, here’s the 10 tools that I have available, plug them into you, and now when you need to use those tools, I will reason and help you pick the right tool to use to achieve the goal.
Dakota Snow• 09:26
It enables the LLM to actually do, like it enables, it gives it visibility, and it kind of helps, I feel, like, bridge that gap. You know, I feel like a lot of people feel like, oh, I’m going to use AI networking, they’re going to start copy and pasting configs into like ChatGPT, which is the wrong way to go about it. That’s not the way to do it. So it enables our network engineers to actually allow the LLMs to communicate safely with the network. Now, here at Itential, you guys also have ways to help manage the changes that LLMs are making, how AI is affecting that. Are you able to kind of speak to that a little bit?
John Capobianco • 10:07
Yeah, so we have this wonderful platform, FlowAI, which went GA today at Cisco Live. We announced the general availability, the Flow AI agents, and they have guardrails. They have role-based access control. You get to tell the agent what it has access to and what is outside of its boundaries. It’s got guardrails built into it, and it has a mission log. So our mission log will show every time the LLM has done some reasoning and its reasoning, and every time the agent called the tool, as well as the tool that it called. So we have a full auditable log of the agent’s activities, which is really great because it lets you really examine the agent’s activity compared to the instructions you gave it.
John Capobianco • 10:51
The other thing that I think is really exciting about Itential is that we have a skill on the Anthropic marketplace. Which means you can install this skill into cloud code, and through natural language, you can build these flow AI agents without any code. So you type in your prompt. The agent does its thing, and within 10 or 15 minutes, the agent is real and executable inside the platform. I’m going to be demonstrating that live here in the Vibops lounge. I think you’re actually going to record the session.
Dakota Snow• 11:24
I am, I am.
John Capobianco • 11:24
So, pay attention to that one because it really is going to show how Itentials flow agents are built, how they’re executed, and all of the guardrails and controls around it.
Dakota Snow• 11:36
So, you know, just to kind of recap the things we’ve talked about, like AI is here, it’s staying, but I feel like it’s not here for our jobs. It’s here to make us better network engineers. And the sooner I think we start realizing that and utilizing that, it’s just another tool in our toolkit. It’s just like all the other tools out there. It enables us to do just so much more. So, the sooner we can start using it and be less scared of it, you know, some fear is okay. You know, it’s good to be, you know, not give it, again, the keys of the kingdom.
Dakota Snow• 12:14
But on the same hand, the sooner we can start utilizing it, the sooner we can advance as engineers and just unlock things that we never, just because we didn’t have time to do these things. Right? Right. You know, it takes so many hours. We are going to be able to do so much more with just our skill levels.
John Capobianco • 12:36
Well, and also network engineers. So, I. My heart goes out to the network engineer that wants to learn network automation but has not. They don’t have the cycles, they don’t have the budget, they don’t have the time or the resources. Programming, learning a programming language, I mean some people go to community college for three years to learn that. That’s not really realistic for a network engineer who’s also trying to get their NP, their IE, keep up on EVPN and SR6 and all the networking stuff. That keeps you busy.
John Capobianco • 13:10
But I think with AI, we can leap to the front of the line. Not to say you shouldn’t learn programming, don’t misunderstand me, but you don’t have to anymore. MCP natural language has democratized and abstracted away so much of that pain of learning how to write classes and functions and proper structured Python code.
Dakota Snow• 13:33
Right, absolutely. You know, before, I got a couple more questions for you, but if people want to go check out some of the stuff you’re doing, like, because you are doing some amazing stuff. You created your own net claw agent, correct? Yes, yes. So that is just mind-blowing that someone could create something like that. If people want to go follow you more, where can they go find some of the stuff you’re doing? Absolutely.
John Capobianco • 13:57
So this QR code should take you to the Vibe Ops Forum, which is a safe, open community for people to come out of the AI closet without being judged that their work is just slop, right? So it’s for people to share the latest news, their open source repos, their projects, their code. So join the community, because community is how we succeed and achieve goals together. It’s, you know, one person alone compared to a whole community of people, you know, the help is there. The other thing I want to mention is that I can be, you know, I’m very open with my time and we’ll try to help you. So reach out on LinkedIn, reach out on X. I’m pretty easy to find.
John Capobianco • 14:38
I have a YouTube channel full of free videos about everything we’ve been talking about. Yeah, just try to take your 1st steps. You know, we can talk about this for hours, but we need to motivate you to just try it. Just try it. The bit in your head will flip from zero to one pretty quickly, and now you yourself will be empowered, like Dakota said, not replaced. You’re going to be empowered by this amazing society-changing technology.
Dakota Snow• 15:11
You know, when you and I 1st started talking, you know, we did a podcast episode not too long ago, which was amazing. That switch flipped for me, and I started viewing AI totally different. Not like, ooh, what does this do? Like, what can it do? What can it do for me? That’s amazing. How can it make my life not only easier, but how can I be more productive?
Dakota Snow• 15:35
How can I be a more valuable engineer? How can I learn more? And that’s another thing: it’s teaching me so much more. I’m diving into rabbit holes I never even thought I would go down. And it’s all because of AI. So, again, I think we really need to dispel this fear and just start using it. The littlest things, and you’ll start realizing that, oh, wait a minute.
Dakota Snow• 16:04
Can it do this? Oh, wait a minute. How about this? And it just opens up Pandora’s box, I really like the spider web. It does. It does. And then, you know, people also fear that using AI too much, we’re going to get too reliant on it and we’re going to lose skills.
Dakota Snow• 16:20
But I argue the opposite: you’re getting more skills, and just how we interact with the network is changing.
John Capobianco • 16:28
I love your message. And the only other thing I would add is that what you should be afraid of is the day when your manager says you have to use AI. Because it’s coming. They’re allocating tokens, they’re tracking token usage. And the last thing you want is to have an uncomfortable conversation with your leadership when they ask you, Are you using AI? And you say, Well, I don’t believe in it, or I’m scared of it. Or, right, it’s a very competitive market right now to get work.
John Capobianco • 16:57
And you know that during the interview process, artificial intelligence is going to come up. So try to at least understand the lingo, the terminology. We’ve just talked about MCP for 10 or 15 minutes. I really, as a heart-to-heart of you watching this at home, right, your future is in your hands. No one’s gonna do it for you, but if you do it yourself and start building Git repos, start building projects, start building MCPs, when those questions come up in the interview, oh sure, I use the Netbox MCP all the time. What do you need? What do you want to know about it?
Dakota Snow• 17:30
Exactly. It’s about standing out from the crowd. Yes. You know, what are you doing differently? You know, everyone now has a certification. Everyone now has the basic knowledge. What differentiates you from everyone else?
Dakota Snow• 17:43
How are you standing out? How are you different? AI is that differentiator, and it’s new enough where you can be the unicorn, much like you have, the role model here for your organization and make yourself irreplaceable. You know, there’s layoffs happening all around. I know. If you are the person, the guiding light for this amazing new technology that is being required in many organizations, you’re going to make yourself, you’re replaceable.
John Capobianco • 18:14
Well, if the agent’s called Dakota’s agent and the whole company is relying on Dakota’s agent, do you not think Dakota is valuable? Right. I think so, right?
Dakota Snow• 18:24
100%. John, thank you so much. Thank you for this. This has been such an amazing time. It’s only day one. I know. And I’m just so excited for what the future of networking has in store for us.
John Capobianco • 18:35
Well, you’re doing a really good job. Your message is positive. You’re helping people. And I’m really proud and thankful for you having me on your platform.
Dakota Snow• 18:43
Awesome. Thank you again, John. Everyone. If you haven’t already, go check out John’s resources. He is, I kid you not, you are like the guiding light right now when it comes to AI and demystifying like this scare and like and just showing us how amazing and cool of a tool it is. So thank you, sir. Thank you for that, everyone.
Dakota Snow• 19:04
I hope you enjoyed this episode. And until next time, keep learning.