Enterprise networks face a fundamental challenge as companies deploy AI strategies: the infrastructure designed for traditional desk-to-cloud traffic patterns cannot handle the demands of agentic AI.
At Cisco Live Amsterdam, we had a chance to talk to Enrico Mercadante, VP Networking Sales EMEA at Cisco. We talked about why Cisco thinks that a network refresh is no longer optional, but rather a strategic imperative for any organization serious about AI transformation.
Why AI changes everything for networks
Agentic AI creates three demands on network infrastructure. First, AI models will increasingly use enterprise data rather than internet data, requiring networks to reach every operational corner of the company. This means connecting IoT devices, sensors, video cameras, and operational technology that may currently be disconnected or isolated.
Second, the volume and pattern of data movement shifts dramatically. Instead of predictable desk-to-cloud traffic, enterprises face high-volume data flows across the entire organization. Think video processing in healthcare and logistics, or real-time sensor data from manufacturing plants. The network becomes the enabler for gathering and moving this data to AI models.
Third, security requirements become exponentially more complex. How do you verify the identity of a sensor connecting to the network? What data should an AI agent be allowed to access? How do you prove the identity of an AI agent itself? These questions put pressure on network security architectures that weren’t designed with AI in mind.
Quantum security
Network refresh decisions made today must account for quantum computing threats arriving in the next five years. Cisco’s newer network devices include quantum-secure boot, quantum-secure management access, and quantum-secure encryption.
This creates a clear dividing line, according to Mercadante: organizations refreshing networks now with backward-looking technology will face another expensive upgrade cycle soon. Those building quantum-safe foundations today can extend their infrastructure lifecycle significantly.
Incremental refresh with strategic planning
The good news for enterprises and the CFOs who need to find the funds to refresh networks is that it doesn’t require shutting down operations for a complete rip-and-replace. Companies can take a phased approach, Mercadante says, but they must start planning now. Many networks already need refreshing due to end-of-support compliance issues, making this an ideal time to reimagine infrastructure for the future.
The key lies in partnership between vendors and customers to create a refresh roadmap, Mercadante argues. This ensures organizations avoid end-of-support vulnerabilities while also avoiding excessive pressure to replace functioning equipment. Strategic planning identifies which infrastructure components need immediate attention and which can wait.
Software-defined networking extends lifecycle
Modern network infrastructure separates hardware foundations from software-defined features. This enables organizations to add capabilities through software updates rather than hardware replacement. Cisco’s Silicon One approach is a good example of how this works. It is possible to update these devices after deployment with new features and functionality.
Also read: Silicon One is the engine under the hood of Cisco’s AI story
Also watch: Cisco Silicon One combines uniform chip design with specific deployments
This architecture extends infrastructure lifecycle significantly. Organizations that build the right foundation can run annual “network refreshes” that consist entirely of software updates. The hardware needs to be robust enough to support future requirements. This includes hardware accelerators for cryptography and processors capable of switching packets at very high bandwiths, but the feature evolution happens in software.
The AI readiness assessment framework
Cisco’s AI Readiness Index measures organizational preparedness across six dimensions: governance, vision, talent, data, and infrastructure (both data center and networking). Infrastructure assessment comes second, not first, Mercadante stresses. Organizations should think freely about AI strategy without worrying about current limitations, then evaluate whether their infrastructure can support that vision.
For network managers, AI readiness tops the refresh priority list. That is, it should top that list, based on what we hear from Mercadante. The question isn’t whether to refresh, but how to embed more security, future-proof critical infrastructure, and prepare for quantum threats while maintaining operations.
Data fabric capabilities through Splunk
Network infrastructure alone can’t solve the AI data challenge. Organizations need clean, reliable, up-to-date data that AI models can leverage effectively. That’s where Cisco’s acquisition of Splunk comes in. That makes it possible to virtualize data across the enterprise. Instead of copying all data to one location, the platform can use runtime data from the edges, making it appear as a single data repository for AI models.
Also read: Cisco brings Splunk to the data, wherever it is
This data fabric approach complements network modernization. Organizations can have excellent network infrastructure, but without a data platform to utilize it effectively, they’re not maximizing their investment. The combination of modern networking and data fabric creates the foundation for successful AI implementation.
Timeline and urgency
At the end of the conversation, we ask Mercadante when companies need to complete this network refresh? The answer to that question depends on how aggressive their AI journey is and how competitive their vertical market, he says. According to him, however, every customer Cisco works with has already started conceptually planning their refresh. The question isn’t whether to modernize, but how quickly to execute the plan.