Cisco reimagines infrastructure for the AI era

From AIOps to agentic operations

Building on this vision, DJ Sampath, senior vice-president of Cisco’s AI software and platform group, elaborated on the capabilities of AI Canvas, which uses generative AI (GenAI) to create dynamic user interfaces, pulling data from across Cisco’s portfolio – including Meraki, ThousandEyes and the recently acquired Splunk – to provide a unified view for IT teams.

The entire infrastructure that powers these datacentres, and AI in general, is going to have to be reimagined, because the scale, proportion, performance and volume is going to be completely different

In a live demonstration, Sampath showed how AI Canvas could take a service ticket, automatically diagnose network issues by correlating data from multiple sources, suggest remediation steps, and allow operators to execute fixes directly. The system is powered by what Sampath dubbed a “breakthrough” deep network model, a purpose-built large language model for networking that he claimed is 20% more precise for troubleshooting and automation.

“We have fundamentally reimagined what it means to be able to do this,” said Sampath, highlighting AI Canvas’s collaborative features that allow multiple team members to work on an issue simultaneously. Patel added that AI Canvas will eventually be extended to Cisco’s entire portfolio, creating a compounding effect as more data sources are integrated.

Supporting this AI-driven software layer is a raft of new hardware. Anurag Dhingra, Cisco’s senior vice-president and general manager for enterprise connectivity and collaboration, announced an architectural approach for campus and branch networks focused on simplifying operations, building scalable devices and infusing security into the network fabric.

For a start, Cisco is bringing Silicon One-based smart switches to the campus network following their introduction in the datacentre. These switches combine the Silicon One network processor with a co-processor capable of running security workloads like Hypershield, effectively turning the switch into a high-speed enforcement point for security policies.

In addition, Cisco has announced two new 8000 series routers to bring Silicon One’s capabilities deeper into edge and access networks of service providers, as well as post-quantum cryptography capabilities to protect against “harvest now, decrypt later” threats, where encrypted traffic is stored today with the aim of breaking it with future quantum computers.

On the security management front, Raj Chopra, Cisco’s senior vice-president and chief product officer for security, detailed a major push to consolidate policy management. All security policies will now be set from a single, authoritative policy engine – Security Cloud Control – and enforced across a distributed footprint that includes new hardware and even third-party devices.

This unified policy engine governs two core architectures: the hybrid mesh firewall for protecting applications and Universal ZTNA (zero-trust network access) for securing user access. The company also launched two new firewall devices, expanded multi-supplier network segmentation, and introduced a fully passwordless authentication option leveraging phishing-resistant technology to secure human and AI agent-based access.

Patel framed the announcements as part of an unprecedented wave of innovation at the company. “In the past 18 months, we probably innovated more than the previous decade combined, and next year will be a multiple of the past 18 months,” he said. “We’re just getting started.”