Thoughts on AI from an Infrastructure Guy

Originally published on LinkedIn on May 27, 2025.

Header image generated with Amazon Nova.

My dad was involved in a head-on collision in a ditch on a hill just 30 seconds from home. What followed was months of surgeries and recovery. When this accident prevented my dad from making the 45 minute commute, he brought home the then coveted team loaner laptop (he may have invented the Work From Home movement). Though modest by today’s standards, not even measuring up to my child’s Kindle Fire, that computer introduced me to…PowerPoint.

Soon, I had my first slide show animating a train moving across the screen. This was the catalyst that began my lifelong interest in technology and continual learning.

“What does this have to do with A.I.?” Glad you asked.

THE CURRENT TECHNOLOGY REVOLUTION

Fast forward 30 years and we are in the middle of the next technological revolution. In fact, this is labeled as the “Digital or Information Revolution” as part of the “Fourth Industrial Revolution” beginning in the late 20th century. It is marked by biotechnology, automation, and Artificial Intelligence. I view AI as its own micro revolution which may turn out to be its very own in the course of time.

Every technological revolution in history has shaped humanity in new ways. Going from hunter-gatherers to agriculture-based required new skills and led to more settlements. The first Industrial Revolution brought mass production with the second one bringing advances in steel and electricity. Once again, this created new jobs, transformed old ones, and led to more urbanization and brand new industries.

And the next ones brought breakthroughs in nuclear power, printing, and computers forever changing the way we live, work, and communicate.

So what’s different this time? In many aspects, nothing. AI is requiring new skills, creating new industries, and will change life as we know it for future generations. There will be those among us who embrace the change early and others who resist to the bitter end. Same as before. However, the AI revolution will differ in several key areas.

UNPRECEDENTED SPEED

The change and adoption rates of AI technology surpasses all previous revolutions. It doesn’t take decades to roll out AI factories or multiple generations to learn how to perfect AI farming. No, this revolution is occurring in real-time before our very eyes. Within a few years, we went from expensive and cumbersome machine learning to “ChatGPT” and now to Agentic AI.

PERSONIFIED INFRASTRUCTURE

Automation and cloud computing introduced me to “infrastructure as code” with templates, software-defined networking, log/trend analysis, and more. AI is different. It can learn and create. Instead of learning a new language or template we now learn prompt engineering, different model characteristics, and how to properly validate output checking for bias and hallucinations. In other words, we need to learn how to interact with not just a new language or feature, but new “personalities” of AI.

ADMIN AT A DISTANCE

Unlike today, where network admins can still access the base configurations, this won’t be optional. AI-Managed Networking (AM-Net) is coming. This will include AI-planned change windows, managed upgrades and deployments, synchronized routing, and more.

Couple this ability with the best practices of Least-Privileged Access and Zero Trust Network Access, and the days of admins inputting statements into Cisco routers such ‘vlan xyz’ without the ‘add’ flag are left behind. Administrators will only plan, instruct and supervise.

INCIDENT AWARENESS

On the post-deployment side, AI will perform synthetic testing and learn using agents across the WAN and LAN. Logs and performance will be managed in real-time leading to truly self-diagnosing and self-healing networks. In fact, this is an idea I’m currently exploring and working on right now (AI-managed packet captures anyone!?). The days of huddling in incident rooms for days on end and performing post-mortems and root cause analysis will be a distant memory.

Instead of incident management and problem solving, we will shift to incident awareness reviewing AI-generated reports and updates.

NEURAL NETWORK REALIZED

And more than all of those, this is the first time I feel we are nearing a true comparison to the human brain and our neural network. For the longest time we have been working with personal computers and data centers. This is akin to a system or two in the human body. We have replicated a specific nerve center in a ‘brain’ communicating with a ‘heart’ to pump and send information through a system.

Additional data centers, servers, and applications bolt on other systems, but they still have never been truly autonomous and integrated. With agentic AI, we now have a controller learning from other agents keeping all systems in check. To take this analogy even further, AI can start creating its own new ‘neural links’ and expand its own capabilities based on experience. All of this will take place across networks and compute. Infrastructure is the foundation.

PREDICTIONS

1. BUSINESS FIRST

So what happens once AI is managing our infrastructure and networking? I am speculating that providers will once again use this as a tool to remove more “undifferentiated heavy lifting” from IT staff and continue to shift the focus to business outcomes. Instead of focusing on specific config templates and routing tables, administrators can review performance and allocate system resource towards priority initiatives.

2. PROVIDER FIRST NETWORKS

Service providers will manage the communication first and isolate customer traffic. Coupled with tech such as IPv6 and data protection instead of ‘walled gardens,’ we won’t need massive private networks. What does remain private to the business may just be a private AI agent that reports to the provider AI. We are already seeing this shift with technologies such as AWS Cloud WAN and edge services like Amazon CloudFront.

It’s only a matter of time before that is controlled primarily by AI.

3. INFRASTRUCTURE CONDUCTORS

As with any technical revolution, the end result is even more human creativity. Infrastructure admins with more time will have the capacity to understand the business, respond more attentively to leadership, and create outcome-based foundations.

Much like an orchestra conductor doesn’t need to tune a violin or write music, an infrastructure conductor will be able to command agents in harmony to deliver a complex global production, a rigid financial march, or a swift small jazz number for a local entrepreneur. Instead of a conductor’s baton they will wield various models crafted for specific purposes. The outcome will be more artistic and nuanced than copied templates.

Services such as Strands Agents have started to bring this to the forefront.

WHERE DO WE GO FROM HERE?

Just like the architect in “The Matrix Reloaded,” technology architects are asking “where do we go from here?” This is a complete paradigm shift new to our generation. It is accelerating at a rate faster than we can comprehend. The potential for it to learn and perform is beyond human capabilities. Never before has a technology been able to surpass human understanding. This will require a complete cultural and societal shift, new understanding, new controls, and “getting it right” the first time.

So, just like my first experience in creating an animated train, we must embrace continual learning and our own curiosity exactly as we are programming AI to do.

If you’d like to continue learning, I suggest not only checking out how AWS can help, but also look at what other companies and groups such as Cisco, Juniper, and Nile are starting to do in this space. Maybe read a blog or pick up a course or two

Note: The opinions shared in this article reflect my own personal views, and not that of AWS.