Cisco’s ‘agentic AI era’ includes AI Canvas for reimagined IT operations

“The way that you should think about us is like the picks and shovels company during the gold rush. We are the infrastructure company that powers AI during the agentic movement,” says Jeetu Patel, Cisco’s president and chief product officer.

Cisco is building out its “AgenticOps” capabilities as the tech giant helps partners and businesses brace for agentic AI’s impact on IT operations.

To that end, AgenticOps, or Cisco’s approach to running modern IT operations that turns real-time telemetry, automation, and domain expertise into intelligent actions, now includes AI Canvas, a generative AI user interface for customer dashboards that lets NetOps, SecOps, and DevOps teams collaborate, optimize operations, while reducing IT strain, according to the company.

“The agentic AI era is upon us,” said DJ Sampath, vice president of product, AI software and platform for Cisco. Agentic AI marks a paradigm shift for businesses and it requires a “reimagining” of operations, he added.

“As we start to think about: How do we implement operational simplicity? It becomes really important for us to recognize that you’re going to have hundreds, thousands, billions of agents that are going to be doing things for you. When you start to think about agents, what really becomes important is being able to operate in an environment where these agents are out there doing the things they are doing. You really need that next level of operations to come together,” he said.

AI Canvas, revealed on Tuesday at Cisco Live 2025, unlocks new agentic AI capabilities, Sampath said.

“It’s going to help you troubleshoot and execute actions across multiple domains of data and multiple architectures. More importantly, different users will be able to collaborate seamlessly on AI Canvas and last, but not least, we’re building all of these on a purpose-built foundation model called the Deep Network model,” he said.

The Deep Network model is one of the most advanced networking LLMs, he said. The domain-specific LLM has been fine-tuned and trained on more than four decades of Cisco expertise, from CCIE-level content to Cisco U. courseware. The data is always being vetted for accuracy and will continuously learn based on telemetry that Cisco will constantly provide, which means this model is going to be very precise, Sampath said.

The Deep Network Model also powers Cisco’s AI Assistant, a natural language interface that can

identify issues, diagnose root causes, and automate workflows, Cisco said.

AgenticOps first and foremost requires a purpose-built, precision model on which to make decisions because AI agents are going to be completing tasks autonomously, Sampath said.

“The second thing that you’re going to need is you’re going to need access to cross domain data to be able to see the whole [environment], to be able to troubleshoot effectively and to be able to solve real problems, you’re going to need to unlock those data silos and connect them all together. And third, you’re going to need oversight, because autonomy without oversight is going to cause a lot of confusion. There’s always going to be humans in the loop,” he said.

For further ease of use and manageability, AI Canvas is part of Cisco’s unified management platform that now brings together management of Meraki and Catalyst devices in one platform that supports any cloud, on-prem, or hybrid deployment that a business chooses.

The agentic AI opportunity

The AI team for Long View Systems, a Calgary, Alberta-based MSP and Cisco Gold partner, is “very excited” about AI Canvas and Cisco’s AgenticOps approach, said Lane Irvine, Long View’s network business solutions director.

For Long View, AgenticOps is the “next step” in what the firm is doing with Retrieval-Augmented Generation (RAG) AI for its customers, Irvine said.

“It’s tying in how do we actually bring the ops side of this into that AI conversation? And how do we actually have that AI engine execute and bring in real-time data for real-time environments. So now, not only are we doing RAG AI with a subset of data and insights, but we’re also then able to bring in real-time insights about the environment,” he said. “From an ops perspective, that really brings together both sides, which is a huge advantage and a huge opportunity for us from an operations and a NetOps and SecOps perspective.”

The agentic era is going to be meaningfully progressive to society in a lot of ways, said Jeetu Patel, Cisco’s president and chief product officer.

“One of the things that gets mostly underestimated and under-hyped in AI, which, if you can believe is actually true in this particular area, is that AI is not just about aggregating information and data that’s there and coming back with clearer answers. It’s actually about creating original insights that didn’t exist in the human corpus of knowledge. And when that starts to happen, you can solve problems that we couldn’t have dreamt solving before, and that’s super exciting,” he said.

But for agentic AI to be useful in the enterprise, fundamental requirements around infrastructure, safety and security, will need to be completely rethought, Patel said.

“Classical infrastructure just won’t be able to deal with the scale and proportion that we’re talking about. Because if you think about it, there’ll be tens of billions of agents that will be conducting work on our behalf, [and] that requires a very different infrastructure [and] capacity in place, and that’s why you’re seeing such a massive level of build out globally of data center capacity as well.”

How is Cisco going to be a part of this era? “The way that you should think about us is like the picks and shovels company during the gold rush. We are the infrastructure company that powers AI during the agentic movement,” Jeetu said.

Cisco AI Canvas will be tested with select customers this fall, according to the company.