At Cisco Systems Inc.‘s annual event, Cisco Live, this week in Las Vegas, it was no surprise that artificial intelligence was the top theme of the show and dominated most of the news and product innovations announced.
Cisco has been successful in riding the AI wave and using it as a growth engine. Over the past year, revenue and profits have grown, and the stock price has doubled. The company has accomplished this by positioning itself as “critical infrastructure for the AI era” and by revamping its entire product line to back that claim.
As customers move from AI pilots to production, the value of the network will continue to grow because AI is inherently a network service. During his Day 1 keynote, Jeetu Patel (pictured), Cisco’s executive vice president and chief product officer, framed the challenge: “Humans click, but agents swarm.” As this shift occurs, spiky, human-led chatbot traffic will give way to consistently higher volumes of network traffic from swarms of autonomous agents running at machine speed, which will stress every part of the technology environment.
Liz Centoni, executive vice president and general manager for applications and customer experience, carried that theme into day two from a different angle: “I’m not here to talk about the future of AI, but what I am here to talk about is the boring problems it’s already fixing for you in your environment.” Her message was that Cisco’s AI story is not just about graphics processing units and glossy demos; it’s about using AI to fix the operational drudgery customers face every day.
Here are five takeaways from Cisco Live that matter for enterprise buyers:
1. Cloud Control is Cisco’s AI-era control plane
If there was one announcement that stood out from the rest in Patel’s keynote, it was Cisco Cloud Control. Patel called it “one of the best pieces of work that I’ve seen a team do,” describing it as “simplicity without losing the sophistication of Cisco.” Cloud Control is an ambitious effort by the company. “Every Cisco product you know will be managed from Cisco Cloud Control, and every new product and acquisition will start there,” he said.
Cloud Control is a unified control plane for all infrastructure domains, with the following benefits:
Cross-domain visibility and action. From a single interface, engineers can view campus and branch networks (Catalyst and Meraki), data center fabrics (Nexus, Application Specific Infrastructure or ACI, Kubernetes and VMware), security controls and collaboration devices. Patel emphasized that every demo on day two — from AI-ready data centers to future-proof workplaces and digital resilience — “will actually originate and start from Cloud Control.”
Agentic operations are the default. Cloud Control is built around AI agents and runbooks rather than static dashboards. In one demo, an AI assistant recognized a mis-cabled switch that created a spanning-tree loop, traced the root cause, proposed a fix and then enabled the operator to promote that remediation to an autonomous action for the next occurrence. Patel described this as “the autonomy dial,” where customers choose when an agent has “earned your trust” before handing over categories of fixes.
Workflows as guardrails. When Patel asked how workflows fit, Anurag Dhingra, senior vice president and general manager of enterprise connectivity and collaboration at Cisco, explained that a workflow is essentially a codified runbook that makes “nondeterministic models behave in a much more deterministic way for predictable outcomes.” If you don’t have a runbook, the agent will generate one from your intent, execute it and save it back to the catalog for future use.
Open harness for agents and data. Patel emphasized that Cisco “can’t be arrogant enough to think that the only technology you’re going to use is Cisco.” Cloud Control was described as a secure agent harness in which Cisco agents, partner agents and customer-built agents run under a common identity and policy system. Thanks to the Splunk-powered Cisco Data Fabric, those agents can reason across network, security, application and AI telemetry in one place.
Strategically, Cisco wants Cloud Control to be the GPS for AI-era information technology operations. Continuously ingesting telemetry, reasoning and proposing the next best action across a complex, hybrid estate. I spoke with several customers at Cisco Live, and they all echoed the same sentiment. They have lived with a sprawl of point tools and disjointed consoles for years and a single, AI-native control plane is highly appealing.
2. Cisco IQ makes the ‘era of guesswork’ obsolete
Day two was kicked off by Centoni and Cisco IQ. Whereas Cloud Control focuses on live operations, Cisco IQ is aimed at lifecycle, risk and support, and it’s arguably the best example of Cisco using AI to turn its own complexity into a customer-facing product.
Centoni began with a reality check: “Spreadsheet after spreadsheet, but you still don’t have confidence in what assets you have in your environment,” she said. “That’s not information, it’s educated guesswork.” In a post-Mythos world, she warned, “human-speed reactive defense is no longer viable.”
Cisco IQ’s value proposition is to wipe out that guesswork:
Real-time, always-current inventory. IQ provides “broad visibility into your Cisco environment, a single, always-current view of all your assets, including hardware, software and cryptographic assets, with no fragmentation or lag.” Centoni urges customers to ask themselves a simple question: Can you answer, right now, “Do you have a complete and accurate inventory of every asset in your environment and its current security status?” If not, Cisco IQ can help illuminate that blind spot.
Post-Mythos vulnerability management. With AI-enabled attacks able to map networks in minutes at machine speed, Centoni framed this as “the Mythos moment” and argued that organizations that don’t move decisively “are not going to get a second chance.” IQ enables customers to see exactly which devices are exposed to a given vulnerability, their lifecycle state, and the most targeted remediation path.
Traction at machine speed. Cisco IQ went GA just weeks before Cisco Live, but Centoni reported that “as of this morning, we have 2,036 customers already onboarded,” far exceeding her own expectation of 800. Customers such as Geodis and GlobalFoundries described IQ as “too much, too fast and too clear,” and then quickly concluded, “This is what it looks like when the system actually knows your environment.”
Support that never starts at zero. Centoni captured the traditional TAC experience in a line every network team recognizes: “You’re spending the first 35 to 40 minutes before a single line of troubleshooting is done on just collecting basic data.” With IQ, she said, “your engineer does not brief TAC. TAC briefs itself.” Topology, config history, prior cases, and logs are pre-populated; Cisco’s AI routes 88% of cases to the right engineer; and resolution starts with context, not interrogation.
Centoni’s closing line is the one many customers will remember: “You’ve always had the map; now you have the GPS. Use it, as the era of guesswork is over.” For buyers, IQ reframes “support” as a data product: a continuously updated view of asset risk and health, integrated with both Cloud Control and Cisco’s human support organization.
3. Cisco Silicon One is the engine of an AI networking supercycle
Patel spent a significant amount of time on silicon and optics, and for good reason. His argument is that AI will drive a massive networking super cycle, and that without rethinking the network from the chip up, enterprises will discover bottlenecks in the places they least expect.
Patel’s diagnosis is that “inference is happening everywhere,” from GPU clusters in data centers to “desk-side computing” with Mac minis hosting hundreds of agents per user. Cisco’s internal study, he said, shows that “on average, every agent generates about 450 more traffic than a human for conducting that same task.” As agents proliferate — his phrase was “trillions of agents that are going to be proliferated everywhere” — the demands on networks will dwarf those of prior waves, such as video.
Cisco’s answer is its Silicon One family, which includes products to scale up, out and across, covering all the critical components needed to build a network that supports an AI factory. Patel noted that Silicon One powers both enterprise gear and hyperscale systems, with different system families (e.g., Cisco 8100 for hyperscalers, Nexus 9300 for enterprises) built on the same silicon. On the campus side, he highlighted the Catalyst 9550 as “the most powerful core switch we’ve ever created,” emphasizing that “if you don’t have a strong backbone, then that becomes the bottleneck of your network” for agentic workloads.
4. Security is being rearchitected for the post-Mythos world
Security was the most urgent thread across both days. Chief Executive Chuck Robbins warned that AI “changes the speed of defense” while “empowering our adversaries at a pace we’ve never seen in our careers.” Patel echoed this view, saying we now live in a “post-Mythos world” where the time from disclosure to exploitation is “a matter of minutes, if not seconds.”
Cisco’s response spans silicon, software and operations:
Live Protect is a bridge, not a crutch. In a live demo, Cisco showed Live Protect on Nexus switches automatically identifying devices affected by a new advisory and deploying shields and runtime compensating controls without reboots or downtime. Tom Gillis, senior vice president and general manager of the infrastructure and security groups, emphasized that “compensating control does not eliminate the need to patch; it’s a bridge between patches,” and Patel underscored that customers still take “about 40 to 45 to patch vulnerabilities.” The design intent is that, over time, customers will “let them go in auto mode” to close exposure windows quickly.
Security is embedded in the fabric. Cisco’s Nexus 9K “smart switches,” powered by Silicon One and on-board data processing units, can host stateful L4 firewall functions directly in the data plane. As was demonstrated, Cloud Control can push a policy that inserts a firewall “between that same AI workload and legacy applications” without external appliances or hairpinned traffic. “No additional appliances, no hairpinning and no complex routing,” Gillis summarized, just “lots more little, tiny firewalls throughout the infrastructure” to contain lateral movement.
Agentic SOC on Splunk. SVP and GM of Splunk, Kamal Hathi, showcased an agentic security operations center in which triage agents discard about 92% of alerts as false positives, guided-response agents orchestrate containment across Cisco and third-party tools, and every agent decision is accompanied by a reasoning log and an evidentiary trail. Patel argued that “in a matter of a few months, a SOC that is not agentic won’t even make any sense.”
5. Cisco wants to help operationalize AI, not just power it
The most important takeaway for business leaders may be that Cisco devoted as much time to AI governance, operations and economics as to raw performance.
A few themes ran through both days of keynotes:
Tokenomics and cost control. In a fireside chat with Advanced Micro Devices Inc. Chief Information Officer Hasmukh Ranjan, Patel walked through a back-of-the-envelope example: If an AI-empowered employee consumes roughly $200 in tokens per week, that’s $10,000 a year. At 40,000 employees, that’s $400 million, and at 90,000 employees, $900 million. Ranjan noted that “that line item never existed a few years back” — and that IT is now “naturally designed to optimize” it. By integrating Splunk and Cloud Control into agent-level observability, Cisco aims to make “tokenomics” a mainstream operational key performance indicator.
Digital twins and trusted autonomy. Control’s digital twin capability creates one-for-one virtual copies of networks, “running the same IOS XE version and the exact same configuration as my network.” Operators can describe tests in natural language, have the AI generate scenarios, and validate changes in the twin before deploying them live. Dhingra positioned this as a way for agents to “earn your trust” and for customers to decide, category by category, when they’re ready for full autonomy.
AI-ready workplaces. Cisco tied its AI story to collaboration and the edge of work with new devices (Room Kit Pro G2, Desk Pro G2, Board Pro G3) that ship with Nvidia chipsets and on-device AI agents such as Director and Note Taker. These devices, combined with Cloud Control and the refreshed campus portfolio, aim to create “future-proof workplaces” that are as agent-ready as data centers.
Final thoughts
Patel closed his keynote by reminding the audience how much has changed in two years: “We used to be a collection of products, he said. “We are now a fully integrated platform, and that’s why we believe that Cisco is the critical infrastructure for the AI era.” Centoni added the operator’s perspective: “Windows open and close,” she said. “The organizations that move decisively inside them get to decide what comes and shape what comes next.”
Cisco has clearly decided what its next act will be. The open question for customers is whether Cloud Control, Cisco IQ and the broader platform deliver enough real-world simplification and trustworthy automation to justify reshaping how they run networks, security and AI operations.
