Cisco’s Post

This convergence between networking, security, and AI resonates strongly with the work I’m doing inside the MIKE‑CORE ecosystem. As agentic systems reshape traffic patterns and accelerate vulnerability discovery, the real challenge becomes maintaining deterministic control over AI behavior — not just at the network edge, but inside the OS layer itself. That’s why I developed the Drift Watchdog Layer (DWL) for AIEWS OS FUSION 4.0: a post‑inference governance engine that evaluates every AI output for normative drift, contextual deviation, workflow integrity, and state consistency. DWL enforces deterministic routing — ALLOW, REVISE, or BLOCK — independent of model weights. Where Cisco is building the connective and defensive fabric for agentic operations at machine scale, DWL focuses on the internal logic layer: ensuring that autonomous agents behave predictably, safely, and in alignment with ecosystem rules. Different layers, same mission: structured intelligence, controlled autonomy, and secure AI operations at scale.

The agentic traffic point is striking — 450% more load reframes the network as the control plane, not just the pipe. Full-stack intelligence at the infrastructure layer — telemetry, Splunk observability, Policy Studio guardrails — is necessary. The gap I still see for regulated enterprises: much of that stack sees flows, agents, and platforms, but not always what text was permitted to reach an external LLM before inference. Cisco AI Defense can shape policy inside the platform. Shadow browser LLMs and ungoverned prompt paths often sit beside that stack entirely. Agentic operations need network-scale defense and execution-scale evidence — block, redact, clear, receipt per prompt — when frontier models are still in the path. The full stack is not only networking plus security plus AI observability. It is also governance at the LLM boundary when agents reason across client, HR, or conduct-sensitive context. Where does Cisco see prompt-level enforcement fitting — inside AI Defense guardrails, or as a distinct control plane in front of the model?

This sounds great! The idea of bringing networking, security, and AI together feels spot on, especially with AI agents creating so much more network traffic. It's impressive how they're using AI to speed up security work, like scanning huge amounts of code much faster. The partnerships with Splunk and NVIDIA, plus work on open standards and quantum networking, show they're thinking ahead. The network is turning into the key foundation for all this AI growth. Looking forward to seeing how it develops!

Operational accountability has to scale with infrastructure intelligence when AI becomes embedded in live service execution. The convergence of networking, security, observability and AI creates the technical capacity for agentic operations, but it also creates a new operational dependency layer. Once autonomous agents begin shaping traffic, triage, policy and service behaviour, the question is no longer only whether the infrastructure can support them. It becomes whether service ownership, decision authority and intervention controls can keep pace with the speed of execution. That is where operational readiness becomes critical. Full-stack intelligence may enable AI to operate at scale, but organisations still need to prove that accountability, control and recovery can scale with it.

Cisco about networking and connectivity and AI infrastructures.. Cisco Live 2026 focused on the convergence of infrastructure and intelligence, defining our unified, full-stack approach. This strategy is essential as agentic AI refactors enterprise traffic patterns; we are responding by building the connectivity and open standards necessary to ensure autonomous agents operate securely and at scale Thank you for sharing

Nice, excellent article. So, I think the next frontier extends beyond networking, security, and AI. It includes governance. And as AI becomes more autonomous, success won’t be determined solely by how intelligently systems reason. It will depend on how effectively we govern the assumptions those systems are allowed to make before they ever reason. Intelligence accelerates decisions. And governance determines whether those decisions deserve to be accelerated.

I’m excited to learn more about Agentic AI for networking and security. Cisco research shows that AI agents can generate up to 450% more network traffic than human-driven work, which creates new scalability and security challenges. As networks continue to evolve, AI-driven security models can help detect vulnerabilities faster, strengthen network protection, and support rapid network growth at scale.

Hello Cisco, Thank you for sharing your courses in your Netacad academy. I am very grateful because I learned a great deal. My recommendation is if you could update the courses with a focus on cybersecurity, such as the Ethical Hacker one and other courses that are very important to continue learning. Thank you, may Cisco keep growing, and I would appreciate it if you also created AI courses focusing on what AI actually is, such as neural networks, convolutional networks, supervised and unsupervised learning, etc., and at the end a section on prompting, since there are already more than enough courses on prompting.

From a manufacturing perspective, the real opportunity begins when AI moves beyond dashboards and starts making reliable operational decisions. That requires secure connectivity, standardized data, and end-to-end visibility across suppliers, factories, and logistics. Technology is only as intelligent as the ecosystem supporting it.