I get extra excited each day as I study one thing new. However, I even have my justifiable share of issues in regards to the future—particularly on the subject of AI and the way it will impression the function of community engineers. Okay… I in all probability have extra than my justifiable share of issues. (That gained’t come as a shock when you’ve been following the previous couple of years of my journey, exploring the “AI FUTURE!!!”)
First off, I wish to be very clear. I’m excited about the way forward for community engineering, community automation, and my place on this great world and neighborhood. In reality, my current weblog, Navigating the AI Era as a CCIE, discusses how superior it’s to be a CCIE proper now.
I usually give attention to the place I see the optimistic prospects. How AI could make our lives and work as community engineers higher.
But in the present day, I wish to speak about one thing that worries me: how the AI future is being mentioned and described. My hope is that by discussing it, we will keep away from the worst attainable dystopian imaginative and prescient of that future. While I like studying books or watching motion pictures about these dystopian futures (a responsible pleasure of mine), I don’t wish to reside in a type of worlds. I’m additionally hoping that you just, my neighborhood, may help me perceive whether or not my concern about the way forward for AI is overblown. So, let’s dive in, we could?
I don’t wish to be an AI babysitter…
There is a phrase that has been exhibiting up in shows, blogs, articles, movies, press releases, authorities documentation, and nearly in every single place else discussing how AI will impression the way forward for work. The phrase refers to an strategy known as “human-in-the-loop.”
So, what is “human-in-the-loop?”
I simply did a Google seek for “‘human in the loop’ ai cisco” and Gemini was useful in giving me this abstract:
Cisco emphasizes “human-in-the-loop” AI, that means integrating human oversight and suggestions into AI techniques to make sure accountability, moral concerns, and dependable decision-making, particularly in areas like safety and knowledge evaluation.
That doesn’t sound dangerous, proper? Here’s one other snippet from a paper I just lately learn on AI and the way forward for job roles:
The extent to which it [Gen AI] can exchange people within the office will rely on the need for human oversight of machine-performed duties.
No doubt you’ve seen or heard comparable descriptions of what it would take to “safely” combine AI into day-to-day duties. Here’s my understanding of why human-in-the-loop comes up time and again in discussions.
It comes down to some factors:
- Using AI presents a “value” companies can NOT ignore. What that worth is can fluctuate, but it surely usually comes down to hurry: AI is just quicker than people.
- AI isn’t at all times proper. And AI can’t be held accountable for errors.
- By having a human log out on the AI work, errors will probably be caught. And in the event that they aren’t, there’s somebody to be held accountable.
I’m NOT saying that the above factors are factually legitimate. In reality, every of these statements on their very own deserves lots of deep consideration and dialogue. But for the sake of this weblog submit, let’s take them as they sit to additional discover my issues a few future the place Hank is a “human in the loop” for AI techniques.
Here’s the issue with “human-in-the-loop”
I like being a community engineer. I like creating community designs to fulfill enterprise calls for. I take pleasure in creating configurations and engineering strong routing protocols. I discover the method of troubleshooting a community concern rewarding.
I’ve spent years of my life studying the talents it takes to DO community engineering. And I nonetheless have a few years forward of me as a community engineer. I even have quite a bit to supply the businesses, networks, and group members I’ll work with sooner or later.
Every description I’ve learn or heard about “human in the loop” locations the human close to or on the finish of “the loop.” An AI instrument is posed an issue, query, or set of knowledge to work on. Then, AI generates its resolution, which is then despatched to a human to assessment, settle for, reject, or make modifications.
When I take into consideration this idea, I can’t assist however conjure up an image of row after row of people spending their days listening for the “ding” of a brand new proposed AI work merchandise, ready for the human to do their factor so the AI can proceed on its “loop,” finishing the work. That simply doesn’t sound like the longer term community engineer I wish to be.
Which will come first: AI or expertise?
There is one thing else I ponder about on this “human in the loop” imaginative and prescient of the longer term. A human community engineer’s skill to determine a mistake made by AI depends on whether or not that community engineer has made that very same mistake prior to now. Or, on the very least, they want sufficient community engineering expertise to note when one thing is mistaken.
As of now, we have now skilled community engineers who can “oversee” AI brokers and determine potential points. Heck, that’s half of what senior community engineers and CCIEs do anyway: help the up-and-coming community engineers on our group by reviewing their work and serving to them study from their errors.
But how will future up-and-coming community engineers achieve the expertise of being a community engineer if they’re merely a cog in “the loop?”
And sure, I’m absolutely conscious that that is an excessive instance and never what folks imply after they say “human in the loop” or “human oversight.” Regardless, it’s vital that we think about this sort of excessive final result now, when the way forward for community engineering is being written. Because I completely suppose there’s a manner this narrative might be rotated—a future imaginative and prescient the place community engineers proceed to be community engineers greater than in title solely.
Let’s flip it round: “AI-in-the-loop”
I suggest that we invert the loop. Make no mistake—synthetic intelligence completely presents worth to community engineers doing community engineering jobs day in and time out. In reality, I exploit it myself. But I exploit AI as a useful resource—like another—at my disposal.
Suppose I’m known as in to troubleshoot an intermittent routing downside at our Internet edge. Using my well-worn community troubleshooting expertise, I collect particulars in regards to the concern, carry out totally different exams, and attempt to replicate it. I examine operational output from the routers and take a look at our community administration techniques. Maybe I ask round, “What changed?”
And if everybody tells me, “Nothing. Nothing changed.” I then ask, “Well, what changed before nothing changed?”
As I do all of this, I leverage many instruments and sources. I’ll seek the advice of our inner documentation in regards to the community. I’ll assessment the current change requests. I’d head over to Cisco.com and seek for error messages or situations. (Well… no, I’ll in all probability go to my favourite search engine and seek for error messages and situations. 
)
It is right here, throughout this a part of my work, the place I’ll deliver AI into “the loop.” Not solely is AI quick, but it surely has been educated on and has immediate entry to all kinds of helpful knowledge that’s related to my work.
AI-in-the-loop: A instrument for community engineers
I could also be struggling to recollect the precise present command to show all the main points in regards to the BGP prefixes discovered by my router. Or I’ll wish to arrange a filtered packet seize and am in search of an instance configuration. Or I’m reviewing tons of of traces of debug messages and will use assist in rapidly discovering the anomalies. These are examples the place AI could make ME a greater, extra environment friendly community engineer.
You see, I’m a community engineer. I’m a fairly respectable community engineer. I’ve typed tens of millions of CLI instructions with my fingers, seen numerous pings drop, configured routing protocols, entry management lists, VPNs, coverage maps, EtherChannels, and so forth and so forth. But I’m nonetheless only a human, not a pc. I’ll not have immediate entry to every little thing buried in my mind, however I do know when the reply is in there. I do know that if I see the proper reply (or one thing shut), I can acknowledge it and get to the answer. It’s the identical purpose an skilled community engineer can resolve a fancy downside with one net search and a look at a discussion board submit or Cisco command reference.
We ought to keep within the driver’s seat. We ought to keep answerable for the networks and the community engineering. We ought to embrace the capabilities of AI to enhance our community engineering work. AI shouldn’t be utilizing us to enhance its community engineering work—we ought to be utilizing AI as a useful resource to turn into simpler community engineers—now and into the longer term.
Really Hank… is that every one AI ought to be?
So, you could be pondering:
Oh, Hank, you good previous boomer community engineer. Get with the instances… AI presents us far more than only a next-generation search engine!
Yes, it completely does—and I’m enthusiastic about lots of the enhancements to the techniques and software program we use each day. Not to say the utterly new techniques and software program which are enabled by AI. Just Cisco’s bulletins within the AI house this previous 12 months excited me about its potential for community engineers.
Just think about what we’ll be capable to do sooner or later. Since the primary community engineer began capturing log knowledge, we’ve acknowledged that it’s almost unattainable for a human engineer to make sense of the flood of knowledge in any well timed trend. Think of all of the outages that might have been prevented if we had been capable of finding the small and early hints buried in counters, NetFlow knowledge, and log particulars. As for safety… wow. There is a lot potential within the safety house to determine and reply quicker.
Embedding AI capabilities into networking merchandise will give us a large increase as community engineers. But this additionally isn’t something all that new. For a few years now, machine studying capabilities have been added and iterated on to reinforce the community assurance options for the campus, WAN, and knowledge heart. They are getting a brand new increase from the GenAI hype and buzz proper now, however most of them aren’t GenAI.
Something is coming to the community engineers’ world that pertains to GenAI that has me very, very excited. Natural Language Interface, or NLI, will quickly be a part of the a lot cherished and lauded Command Line Interface (CLI) and the slightly-bummed-it-isn’t-the-new-kid-on-the-block-anymore Application Programming Interface (API) as strategies community engineers work together with the gadgets and techniques we handle. And that will probably be superior. Truly, a sport changer.
Yes, a part of turning into a community engineer is studying all the precise instructions required to make the community work. When community engineers collect collectively and share struggle tales, somebody will at all times complain (lovingly) about the way it is unnecessary that it’s “ip ospf authentication-key” however “ip authentication mode eigrp,” and why can’t they only be the identical?! And we’ll giggle and giggle and giggle.
But let’s be sincere. It isn’t memorizing particular command line syntax that makes us community engineers. It is realizing how, why, and when we have to configure authentication for our routing protocol that’s vital. Won’t we be a lot happier once we can merely inform our router:
“Enable authentication for EIGRP and OSPF on all interfaces. EIGRP should use md5 with key-chain 5, and OSPF needs to use plaintext because of the legacy device we are connected to.”
Sure, some community engineers will grumble and say issues like “back in my day.” But I do know I’ll be happier for all of it.
So what now?
So what now, you ask? Well, I wish to hear what you all suppose. Don’t be shy. If you suppose I’m overreacting, please inform me. If you share my issues, let me know I’m not alone. What excites you about the way forward for community engineering with an AI assistant in your pocket? Are there some duties you possibly can’t await AI to take over for you? Leave a remark under to let me know your ideas!
In the meantime, listed below are some ideas for wonderful locations to study extra about AI and begin constructing expertise. Because there’s one factor I’m completely positive of… AI is coming, and we gotta be prepared for it.
- Spend about 45 minutes Understanding AI and LLMs as a Network Engineer with this nice tutorial by Kareem Iskander.
- Invest extra time on this wonderful Network Academy course, Introduction to Modern AI, with my new favourite teacher, Eddy Shyu. (Don’t let the truth that it’s on Network Academy scare you away. It is implausible for anybody trying to get a stable basis in AI.)
- Dive in deep and “Rev Up” your recertification journey (34 Continuing Education credit!) with AI Solutions on Cisco Infrastructure Essentials. Free in Cisco U. till April 26, 2025, and with content material and movies from 5xCCIE (and my hero) Ahmed Moftah.
Sign up for Cisco U. | Join the Cisco Learning Network.
Follow Cisco Learning & Certifications
X | Threads | Facebook | LinkedIn | Instagram | YouTube
Use #CiscoU and #CiscoCert to affix the dialog.
Share: