Cisco Cloud Control: A Shared Control Room Among Operators and AI Agents
Cisco Cloud Control: A Shared Control Room Among Operators and AI Agents
The transition from chatbots to agentic AI is already having a concrete impact on corporate networks: the traffic generated by software agents far exceeds that produced by humans, and according to Cisco's estimates, it is expected to triple over the next three years as agents proliferate to individual computers. This is the premise with which the company built the announcements at Cisco Live, the technology event that took place in the United States at the beginning of June, later discussed in a meeting with the Italian press to frame them within the overall strategy. The underlying thesis is that the network is the foundation on which large AI systems rest: without a performing infrastructure capable of handling the load of agents, those systems do not function.
At the center of the picture is Cisco Cloud Control, which the company presented as the main announcement of Cisco Live and as the basis of its operational model, dubbed AgenticOps. The idea is to enable human operators and AI agents to work together in the management of critical IT infrastructures: a single access point and a unified view that combines networking, security, computing resources, observability, and collaboration, with control always remaining in the hands of humans.
"AI agents continuously reason and act at software speeds, fundamentally changing the way we scale, manage, and protect critical infrastructures," declared Jeetu Patel, President and Chief Product Officer of Cisco, from the stage at Cisco Live in the United States, describing Cloud Control as a command center where human teams and agents operate within the same environment, with the same information, and with humans always at the helm.
A Platform, Not a Product
In the Italian meeting, Fabrizio Gergely, Cloud & AI Sales Leader of Cisco South Europe, framed the company's direction: a deep vertical integration, from chips to optics to software, but accompanied by an open platform towards the rest of the ecosystem to avoid any form of lock-in. The watchword is to rethink how customers use technologies, shifting the focus from individual devices to the operational model.
On this ground, Luca De Fazio, Account Executive Enterprise Networking of Cisco Italia, insisted on one point: Cloud Control is not a product but a platform. The declared goal is to transform the ambition of agentic AI, which often remains stuck in isolated pilot projects and separate interfaces, into something usable in a production environment. The platform rests on three pillars. The first is a multi-domain telemetry, where data from Cisco products flow into a single data lake enriched with partner integrations and customer customizations. The second is a set of specialized models that work on that data, including the Deep Network Model, trained on proprietary documentation and 40 years of the company's operational network data. The third consists of agents that, based on that foundation, follow a structured path from problem identification to solution verification.
Above this base, Cisco has built distinct working environments. AI Canvas is the shared space where operators and agents address problems together while maintaining the same context, while Cloud Control Studio allows for the creation of custom agents and applications, with native connectors or through the open standard MCP, then to be published on a marketplace that already counts over 50 platforms and third-party tools. To support the model, the company presents data from its own experience, being its own first customer: the average time to resolve a problem has reportedly decreased from seven to two days. The control remains with human operators, called to make decisions, while agents handle repetitive tasks.
Security Moves Inside the Network
The second block of announcements concerns security, illustrated by Renzo Ghizzoni, Country Leader Sales Security of Cisco Italia. The scenario has changed with the arrival of frontier models capable of detecting vulnerabilities and chaining them into attacks at unprecedented speeds: the window between discovering a flaw and exploiting it has shrunk from weeks to a few minutes. Cisco claims to have scanned approximately 1.8 billion lines of code in eight weeks, a task that would have taken eight years with traditional methods. The response comes from Live Protect, a protection that operates during execution without reboots or interruptions, already available on the N9000 series switches and forthcoming on other products, and from a Zero Trust approach applied to agents, with control not only of access but also of behaviors.
The strategy operates on two fronts: protecting agents from the outside world and protecting the world from agents, which can also become tools for attack. This includes AI Defense, which validates models and imposes guardrails extending to MCP servers and agent frameworks, and the planned integration, set for July, of agent protection capabilities directly into the Cisco Secure Client on user devices. On the identity front of non-human entities, Cisco confirmed its intention to acquire Astrix Security, while the Hybrid Mesh Firewall aims to limit lateral movements in the event of a breach through fine segmentation.
Data analysis is handled by Splunk, now integrated into the Cisco offer through the Cisco Data Fabric and an increasingly agentic SOC, capable of detecting and responding to attacks at machine speed, as explained by Michele Apa, head of the Splunk Solution Engineer team for Italy. The same framework, applied to various types of data, was demonstrated in a demo with a Formula 1 simulator at the Cisco Innovation Center in Milan, where the presented solutions are available for customers and partners.
Long-Term Vision: From Quantum to Sovereignty
A third thread looks further ahead, to quantum computing. Cisco estimates that the first large-scale quantum computers could arrive within three to four years and warns against so-called harvest now, decrypt later attacks, where encrypted data is collected today to be decrypted tomorrow. The company aims to bring quantum-safe communications across most of its core portfolio by December 2026 and introduces Quantum Ready Assessments, available through Cisco IQ, to identify the most exposed assets and define intervention priorities. Cisco IQ, integrated into Cloud Control, gathers 40 years of the company’s know-how and will also offer on-premise installation options to meet data sovereignty needs.
The positioning that Cisco claims is that of a company capable of covering the entire infrastructure, from campus to data center to security and collaboration, and to do so with a vertical platform on its own technologies but open towards the rest of the ecosystem. When questioned by Edge9 about managing non-exclusively Cisco environments, De Fazio acknowledged that integration between different products remains the responsibility of the customer and represents a working area for partner system integrators, made possible by the exposure of MCP servers. It remains to distinguish what is already available from what is still promised: Cisco Cloud Control is currently in Controlled Availability in the United States, with global availability to follow, and several security functionalities and Cisco IQ are expected in the coming months. The bet is that the openness of the platform, more than the individual feature, will be the lever with which the company intends to oversee infrastructure management in the era of agents.