Presented by Cisco
Few enterprise technology shifts are more significant than the rise of AI agents. This means a leap in digital headcount that will force every CIO to revisit infrastructure once taken for granted: the network.
Matt Marshall, editor in chief of VentureBeat, welcomed Anurag Dhingra, SVP and GM of enterprise connectivity and collaboration at Cisco, to the latest VentureBeat in Conversation. They talked through where networks fall short, what AI-ready really means and how companies can turn connectivity into a competitive edge.
“When agents — whether they’re embodied AI or software agents — become ubiquitous, it’s going to feel like the workforce went through a step function increase,” Dhingra said. “Of course, network traffic is going to go through the roof as a result. So, you have to be very sure you have the right infrastructure to be ready for that AI future. Unfortunately, not many organizations today do just yet, despite the fact that this is not an optional technology going forward.”
New architecture for a new era
“Connectivity from the office environment to the internet had to be reimagined and rewired for the transition to cloud and SaaS, and I believe we’re at the cusp of a similar technology transition,” Dhingra says. “It’s not just about agentic AI. It’s around taking a step back to start thinking about what the right architecture for this new era — AI-ready and secure — looks like.”
There are three elements to consider: first is ensuring you have scalable and secure devices that are ready for all the traffic and the security exposure that’s expected in this new world. Second is about fusing security into the fabric of the network, so all of the new threats that are surfacing are manageable. And then finally, AgenticOps, or leveraging AI to help simplify operations, which is a boon to IT departments that are struggling with shortages of skilled people.
Beyond the question of wired vs. wireless
Cisco has delivered on its promise to converge its full portfolio across switching, routing, wireless, and even industrial networking, into a single, unified platform. Through this platform – manageable on-premises, in the cloud, or a mix of both – customers benefit from end-to-end network assurance, rich telemetry and insights — even into unowned infrastructure online that connects you to cloud services. All of that is accessible through a conversational agentic AI interface that helps users make sense of the data and can also take actions on the user’s behalf.
The Cisco C9350 and C9610 Smart Switches enable a more efficient and extensible architecture. These new devices can run networking and security processes in parallel, without impacting the performance of the switch, by integrating services directly with the network, rather than bolting them on top. By combining their own Cisco Silicon One ASICs with this co-processing capability, they’re not only weaving security directly into the fabric of the network, they’re making it easy for IT teams to scale and adapt networks as business needs change, without adding expensive new compute or new hardware. And that’s where cost savings emerge: hardware consolidation, reduced power consumption, and operational simplicity.
This approach builds on the journey Cisco began with smart switches in the data center, where security was directly embedded into the switching layer with AI-native, hardware-accelerated architectures like Hypershield. By integrating security into the fabric itself, smart switches reduce the need for additional appliances and enable data center operators to create micro perimeters around every service that makes up a workload.
But now that traffic is increasingly impacting the campus network, Cisco is extending this same concept to campus environments. The switches incorporate two processing engines: a high-performance network processor for stable data transfer, and a network services co-processor for agile security processing, with traffic steered between them intelligently, to ensure performance remains fast and stable.
Boosting network security
With new attacks emerging that target both network infrastructure and encrypted data flowing across the switch, security for the campus and branch networks requires a multilayered approach. It starts with securing the device itself, and then one layer up is connectivity, and the top layer is securing users, applications, and devices.
“When you look at the core of this layer stack, how do we build trustworthy systems?” Dhingra says. “This is where our custom ASIC, Silicon One, has some amazing security capabilities. We can provide tamper-proof devices that are ready for post-quantum cryptography. One layer up, you can also encrypt data flows with post-quantum crypto. On top, where capabilities like segmentation and universal zero trust network access sit, you can define policies in one place, and you push those policies out to every network element.”
Each device has a context that carries across user identity, device identity and machine identity. With a policy in that context, each device can apply the right type of security policy so that enforcement becomes truly distributed, but policy is managed in one place. It’s a multilayered approach to address new risks that are emerging.
Security policies are automatically updated to the right enforcement points to keep security posture up to date. Plus, organizations can make policy lifecycle management work at scale by using self-qualifying policy updates before deployment, and extend consistent policy enforcement across multiple domains. Because policies can be managed across a library of enforcement points in the cloud, on-prem and on traditional next-gen firewalls, customers now have a single management system with Cisco Hybrid Mesh Firewall.
The solution, seamlessly integrated into existing processes, supports common and separate workflows for NetOps, SecOps or NetSecOps teams using a single solution to maintain connectivity and security.
AI assistants for the network
Today many developers are using AI to write code, and in some cases the agents writing that code are completely autonomous — but in all cases, a human evaluates that code before it’s put into production. Cisco’s new agentic assistants for network management take inspiration from those workflows, by offering configuration change suggestions based on what they’re observing in the network, which are reviewed before they’re deployed to put guardrails against the possibility that these models will hallucinate.
Cisco has also introduced a collaborative workspace for NetOps, SecOps and DevOps teams with AI Canvas.
“This is how we’re going to create a next step forward in AgenticOps with humans in the loop, simplifying the workflow to deliver amazing experiences to people,” Dhingra explains.
An admin who’s using an AI assistant with a conversational model to troubleshoot a network can pull in another coworker into a collaborative canvas that incorporates all the charts and graphs being dynamically generated by the AI assistant and invite them to take a look and help unknot any issues.
Ensuring reliability and service across the whole network
“We think of assurance as a promise,” Dhingra says. “It’s a promise to deliver amazing experiences to people who are connecting to the network. That’s why we build capabilities that are not just about monitoring what’s going on, but proactively and predictively taking steps to not have any incidents or any outages to begin with. That’s the holy grail, the journey we’re on.”
In the meantime, it’s about getting visibility across every piece in your infrastruct6ure, and every piece that is outside of your infrastructure. Cisco takes a layered approach, with rich telemetry coming out of devices across the switching, routing and wireless domains, which can be stitched together to make sense of the data. Layered on top are synthetic tests, which are enabled by ThousandEyes.
The product offers a single time-correlated view of critical metrics, based on all the networks and services that make up the user experience. Its cross-correlation algorithms produce cross-layer telemetry as well as interactive visuals to make it easy to plan service rollouts, isolate problems, take action, and resolve issues faster.
“We’re taking a step forward in the ability to apply AI to simplify operations,” Dhingra added. “It’s really about operations at scale. I’m very excited about the new capabilities both within Cisco and as agentic AI evolves.”
Sponsored articles are content produced by a company that is either paying for the post or has a business relationship with VentureBeat, and they’re always clearly marked. For more information, contact
