82 per cent of cyberattacks are now AI-enabled. 14 per cent of organisations feel prepared for them. If that stat scared you, you’re not the only one.
I recently spent a week at Cisco Live, talking with Cisco execs and pouring through product updates to understand what the emergence of AI means for today’s infrastructure investments that will keep our organisations running into the 2030s.
Cisco Live 2025 showcased the shift from promise to reality. The big ideas we’ve been talking about are now real, deployable technology. In a market full of AI buzzwords and “cloud-native” claims, it was refreshing to see actual delivery and products you can deploy today.
So, what rose to the top?
I remember when security meant Cisco PIX firewalls and maybe some ACLs on the core router. Yes, I’m that old. Those were simpler, but dangerously inadequate times, and we didn’t fully appreciate that until hackers started moving laterally through our networks like they owned the place.
Cisco Hypershield is the first true architectural shift that I’ve seen in decades.
For those interested, here’s the technical foundation: Cisco acquired Isovalent, giving them kernel-level packet control through eBPF (extended Berkeley Packet Filter), a technology for efficient packet processing. Cisco have now built (“Smart”) switches with Neural and Data Processing Units (NPU / DPU) that deliver line-rate firewalls in every single physical port.
Security isn’t bolted onto the network, it’s embedded at the silicon level.
However, the real innovation is the enforcement. Hypershield creates traffic policies enforced in the DPU, extends filtering inside virtual server operating systems, and reaches into application containers. Security isn’t a chokepoint anymore, it’s a fabric that follows workloads whether they’re physical, virtual or containerised.
The old world of adding security to create a “secure network” is gone. Today, we build networks where security is a base feature. This is what Gartner calls “unified firewall mesh,” but Cisco’s taken it further than the analyst term suggests. When your Kubernetes container spins up, security policy is already there. When traffic moves between pods or services within the Kubernetes cluster, security policy is enforced restricting lateral movement. When packets hit a physical port, enforcement happens before they’re switched.
Let’s address the threat nobody wants to think about: post-quantum cryptography.
If you’re new to the term, I’ll explain.The encryption used today protects everything from WhatsApp messages to the global financial system. Quantum computers, once big enough, will crack today’s encryption without breaking a sweat. The only solution is to use new types of encryption that quantum computers can’t crack.
Now, while quantum computers are real, they’re not quite up to the task of cracking your banking traffic, yet. So why does this matter now?
Network switches have a 5-plus year lifecycle. The equipment you’re buying now will still be in production well after quantum computing cracks today’s encryption. Nation-state actors are already storing today’s encrypted data, expecting to exploit it once they can decrypt it in the future, a strategy known as “harvest now, decrypt later”.
For government, defence, financial services, or anyone with data that stays sensitive for years, this isn’t futuristic speculation, it’s risk management for today’s investments.
The investment conversation is pretty straightforward: do you want rip out perfectly functional switches in a couple years because they can’t protect against new quantum attacks, or should you buy now to ensure you’re protected against a very real threat?
Today’s spend should be on hardware that’s already quantum-resistant, for example, Cisco Smart Switches.
Here’s another stat for you: 46 per cent of security teams spend more time managing their tools than managing threats. That’s not a security problem, it’s an operations problem. Tool sprawl is killing productivity across the board with multiple management consoles, inconsistent interfaces and endless device-by-device configuration crushing the soul of NetOps and SecOps teams around the country. It’s death by a thousand clicks.
Cisco launched Hyperfabric last year for their 6000-series Switches. Think Meraki-style cloud management for data centre switching with GUI-based design, auto-configuration on plug-in, and visual wiring validation that shows you cable problems before they cause outages.
This year’s expansion is what makes it actually useful at scale. Hyperfabric now supports Nexus 9000 switches, which is massive. The Nexus 9000 is one of most broadly deployed data centre switches, with many of our customers relying on it. Now companies can upgrade their operating model without ripping and replacing kit that still has years of life left in it.
The announcement means we’re not choosing between “easy but limited” or “powerful but complex” anymore. We finally get both.
Cisco have also closed the gap between their cloud networking (previously “Meraki”) and traditional Catalyst switches.
Historically, you had to choose between Meraki simplicity or Catalyst power. You couldn’t have enterprise-grade resilience and micro-segmentation with cloud management. That trade off is now gone.
Catalyst switches now connect to cloud management without sacrificing the capabilities that made them enterprise workhorses. Teams skilled in CLI can gradually adopt cloud automation instead of being forced into a sudden overhaul.
The plethora of Hypershield, Hyperfabric and Cloud Networking announcements all point to an accelerated shift from device-centric to policy-centric operations.
You define intent once (“finance need this security” or “IoT devices get this level of segmentation”) and push it out, everywhere. This is how we combat tool sprawl. Unified management. Consistent policy. Automated enforcement. Teams stop context-switching between platforms and start managing their networks, not their tools.
Everyone’s talking about AI. Much of it is marketing noise, but some of what I saw last week blew me away.
Chris Day, a Principal Consultant here at Data#3, showed how he uses an AI tool to design and configure Cisco Splunk services. Everything from data extraction and data validation to enterprise dashboards. His demo wasn’t just AI-assisted, it was AI-led. Showcasing self-replicating specialised agents that handed off tasks between one another, mimicking the way a human engineering team collaborate but in a fraction of the time.
This is a force multiplier story happening in real time, not some ‘coming soon’ trailer.
Chris wasn’t replaced, he was elevated. He’s doing strategy and business process mapping while agents do the work. AI agents don’t get bored or make typos at 2am, and they help us overcome of our industry’s biggest challenges like the software skills gap.
Cisco’s Unified Edge rounds out their AI strategy. It’s a modular chassis that combines networking, compute and storage in a single device designed for branch offices, retail stores, and remote sites. The key detail being NVIDIA chips embedded for local AI workload execution. Think about what this enables.
Real-time customer behavioural analysis in retail stores without backhauling video to a data centre. Predictive maintenance at manufacturing sites processing sensor data locally. Video analytics at branch offices that don’t overwhelm your WAN links.
And because it’s a unified chassis, you’re not managing separate SD-WAN, compute, infrastructure, and storage arrays at every remote location. It’s all in one box with centralised policy management through Cisco Intersight.
The new Webex Contact Centre AI virtual agents showcase another angle: AI handling the 80 to 90 per cent of routine interactions so humans can focus on complex cases.
These aren’t the robotic IVR (Interactive Voice Response) systems everyone hates. They’re natural, conversational AI that understands context. Customers get better experiences because they’re not stuck in support hell, and human agents get better jobs because they’re solving interesting problems instead of answering “what’s my account balance?” for the thousandth time.
The pattern across all these examples is the same: AI amplifying human capability, not replacing it. Strategic direction stays human. Complex problem-solving stays human, while repetitive execution and high-volume routine work shifts to AI. That’s the actual promise of AI in enterprise infrastructure. There aren’t sentient networks, but intelligent systems that make your teams more effective.
Yes, most of the week was about the tech, but it would be remiss of me if I didn’t include a comment about the emerging members of our industry that I met at the National Industry Innovation Network (NIIN) booth.
I spent a few minutes watching university students demonstrate AI-accelerated 3D visualisation of cybersecurity threats. A global map overlay with real-time threat intelligence rendered as immersive spatial data, making pattern recognition intuitive instead of forcing analysts to parse endless log files.
Two things struck me.
First, the immediate value: Humans are visual creatures, and we need better ways to see threats. These students built it using gaming engines and spatial computing. Non-traditional approaches to B2B problems that are most effective than traditional dashboards.
Second, the strategic signal: this is where our future engineers are coming from, and they’re not thinking in terms of CLI and log files, and they’re not using traditional tools. Years ago, I would have created training plans for young engineers and lamented the cost to train new members of the team to become effective, now I think I may need to create training plans for new engineers to educate the existing team!
When you evaluate infrastructure today, stop asking “which switch has better specs?”, and start asking “what are the battles I’ll be fighting in 2030?”
AI-enabled attacks are already here, influencing 82 per cent of security incidents. Quantum threats are here, even if their impact won’t be felt yet. Your workforce is distributed. Your applications live in the cloud. Tool sprawl is killing productivity.
Every technology decision locks you into into a trajectory that’s expensive to change. The network-as-platform approach demonstrated at Cisco Live isn’t optional, it’s the only way forward that makes operational sense. Embedded security. Policy-driven management. AI amplifying human capability. Infrastructure that understands application context.
Security isn’t something you add on to a network. Either you have a network (with embedded security), or you have an insecure network. You can choose to buy “networking gear” that passes packets, or you can invest in intelligent infrastructure that adapts, protects and scales with your business over the next five-plus years.
The question I walked away with is simple: is 2026 the year organisations stop thinking of networks as a collection of switches and bolt-on technologies and start treating them as the intelligent fabric they need to be?
I’d be interested to hear what’s top of mind for your organisation. Reach out to me via linkedin
Information provided within this form will be handled in accordance with our privacy statement.