Tech bellwether Cisco is weaving artificial intelligence (AI) across its collaboration, networking, and security portfolio to make enterprise networks "agentic," self-diagnosing, and easier to operate at scale, according to Herns Hermida, solutions architect at Cisco Philippines.

In an interview at the sidelines of the 8 Tech Summit in Shangri-la Boracay, Hermida outlined how Cisco's AI strategy spans Webex collaboration features, AI copilots for network and security admins, and a coming data fabric powered by Splunk (now part of Cisco) to correlate "machine data" across multi-vendor stacks for real-time operations and prediction.

Collaboration: AI that hears, sees, and summarizes

Hermida said Cisco has invested in technologies that sharpen the Webex experience: industry-rated noise suppression (think dogs barking and street sounds), dynamic background control that auto-blurs when a user stands up "for privacy," real-time translation, speech-to-text, and automated meeting summaries with action items.

"It takes audio cues and turns them into action items," he noted.

Networking: a domain-trained AI assistant and 'agentic ops'

On the network side, Cisco has embedded an AI assistant into management platforms like Catalyst Center and the Meraki dashboard.

The assistant is backed by a generative model trained on decades of Cisco product knowledge and TAC (Technical Assistance Center) cases.

Admins can ask, "What's wrong with my network?" or "How do I reconfigure this router?" and receive guided troubleshooting, he said.

Hermida also highlighted "agentic operations" via an AI Canvas that pulls signals from across the environment to build root-cause analyses and executive summaries automatically -- then recommends (and helps execute) fixes.

Cisco's security stack now includes an AI assistant to review and rationalize firewall rules and to suggest changes. AI also prioritizes relevant alerts across pipelines, "filtering out the noise" for SOC teams, Hermida said.

Behind the scenes is Talos, Cisco's 24/7 global threat intelligence and incident-response organization. Talos tracks zero-days, issues advisories, and helps customers investigate and recover from breaches.

"If the logs are there, it's easier," Hermida said of restorations, adding that Talos leverages experience from global incidents to guide forensics and remediation.

Getting AI-ready: resilience first, then capacity and governance

Before enterprises chase AI use cases, Hermida urged them to harden the basics:

Architecturally, customers increasingly "train in the cloud, infer on-prem," which drives stringent requirements for bandwidth, latency, and data-loss prevention.

To ease data center protection, Hermida said Cisco is introducing smart switches that embed firewall capabilities directly in the switch fabric -- shrinking footprints and enabling consistent Zero-Trust Network Access (ZTNA) across sites and even between agentic AI applications.

On compute, Cisco has partnered with NVIDIA to deliver GPU-ready servers "when you need them," while emphasizing that not all AI workloads require accelerators.

Inside Cisco: AI for employees, trained on Cisco's corpus

Internally, Cisco uses commercial and in-house GenAI tools trained on Cisco data. Employees can query an internal LLM for everything from HR processes to "the best configuration for this model router," Hermida said, with accuracy and productivity as the goals.

What's next: agentic networks and the machine-data frontier

Over the next three to five years, Hermida expects AI to make network operations agentic by default: admins will describe an outcome in natural language, and coordinated agents will diagnose, act, and monitor results "in minutes," enabling fewer people to run much larger networks.

A major pillar is the Cisco Data Fabric vision with Splunk, which correlates telemetry from servers, networks, firewalls, apps, and more -- then layers AI to predict demand spikes, surface business impact (e.g., real-time revenue loss during outages), and trigger just-in-time scaling.

"The next frontier of AI is machine data," Hermida said. "Correlate across different vendors, then add prediction so you can act before things break."