By Jason Gintert, CTO

From Dial-Up to AI-Driven Networks

After nearly three decades in networking, I’ve been fortunate to see more “revolutionary” technologies than I can count. Many fade as fast as they appear. AI is different in that it has matured past theory into production ready systems that solve real problems.

Take HPE Aruba Networking’s Central, which brings AI insights, workflow automation and enterprise-grade security into one cloud dashboard, simplifying management across campus, branch, remote, data center, and IoT networks. Or HPE Juniper Networking’s AI-native Mist platform—with its Marvis AI engine, customers have reduced network-related trouble tickets by up to 90%, time spent managing networks by up to 80% and OpEx costs by as much as 85%.

These are field-tested results from production environments, not lab demos.

“Customers have reduced network-related trouble tickets by up to 90%, time spent managing networks by up to 80% and OpEx costs by as much as 85%”

What We’re Building at Laketec

We believe the best way to serve clients is to prove it works on ourselves first. That’s why our engineering team is developing AI-powered network management integration using Model Context Protocol (MCP), an open standard that enables AI assistants to securely connect to your infrastructure tools.

“The best way to serve clients is to prove it works on ourselves first”

Here’s what this means in practice. Instead of juggling SSH sessions, web dashboards, and CLI commands, you ask questions in plain English:

• “Show me all APs that went offline in the last 24 hours at the Dallas site”
• “Compare the device inventory between Auvik and Aruba Central and flag discrepancies”
• “Which switches have CPU utilization over 80% across all customer sites?”

The AI queries multiple platforms simultaneously, correlates the data, and presents unified results. We’re currently running this in production for read-only operations, gathering insights without touching configs. Configuration management comes next, after we’ve battle-tested the security model.

Why the Timing Matters

The pace of improvement is staggering. Microsoft 365 Copilot went from mediocre to genuinely useful in just months. HPE has spent over a decade advancing AI Operations (AIOps), accelerating the journey toward self-driving networks. The inflection point is here.

For IT leaders, this shift delivers measurable outcomes: fewer tickets, faster resolution, lower OpEx and most importantly, more resilient infrastructure that prevents issues before users notice.

Your Path Forward

If you’re skeptical, I get it. AI is not a silver bullet. But it’s too valuable to ignore. Here’s how to start with networking AI specifically:

Today (Using What You Have):

• Enable AI features in your existing platforms (Aruba Central’s AI Insights, Juniper Mist’s Marvis, Meraki’s ML features)
• Start with anomaly detection and predictive failure alerts. Low risk, high reward
• Test natural language queries if your platform supports them: “Show me the worst performing SSIDs this week”

Next Quarter (Expanding Capabilities):

• Integrate AI with your ticketing system to auto-classify and route network issues
• Explore AI-powered capacity planning for wireless and switching infrastructure
• For sensitive environments, explore self-hosted models like Llama or oss-gpt that run completely offline

Measure Everything:Track metrics before and after: mean time to resolution, tickets per device, false positive rates. If a tool doesn’t move the needle within 90 days, drop it.

“This is one of the most significant shifts in technology since the internet itself”

The Bottom Line

This is one of the most significant shifts in technology since the internet itself. Organizations that learn to wield AI effectively will have an edge in efficiency, reliability and cost control. Those who hesitate risk falling behind.

At Laketec, we’re proving the value by building these AI capabilities internally today. Our MCP integration is live in our labs. By Q2 2026, we’ll be rolling out managed AI network assistance.

The tools are here. The results are real. The only question is: will you take advantage of them, or let your competition get there first?