Video Summary
Enterprises are moving AI inferencing to the edge to reduce latency for robotics and real‑time systems.
Agentic AI lets teams delegate routine network tasks to software agents while humans remain in oversight roles.
Cisco Unified Edge bundles compute, networking, and security to simplify on‑premise AI inferencing and remote management.
AI Defense and multi‑layer security (ASICs to app level) address model/data risks and emerging quantum‑era threats.
Network engineers should adopt AI incrementally to boost productivity and focus on critical thinking and continuous learning.
Agentic AI refers to software agents that can perform tasks autonomously; in networking they can handle routine operations, detect anomalies, and recommend or execute changes under human supervision, enabling IT teams to scale despite skill shortages.
Local (edge) AI reduces latency for real‑time and embodied applications—like robots on an assembly line—improving reliability and performance compared with running inference in distant data centers.
Unified Edge combines compute (optionally with GPUs), networking, and security in a single chassis to simplify deployment and remote management of on‑premise inferencing at retail, factories, and other sites with limited IT staff.
Cisco offers AI Defense for visibility and protection against model/data attacks (like prompt injection and data leakage) and is building multi‑layer security including tamper protections and post‑quantum encryption to counter 'harvest now, decrypt later' risks.
Start by using AI to augment current tasks and gain 20–40% productivity boosts, then reimagine workflows; focus on critical thinking, continuous learning, and verifying AI outputs rather than blindly trusting agents.
"More and more, you'll see robots building things powered by an AI model, and when these models run far away in a data center, latency affects the assembly line."
There is a growing trend among enterprises to deploy local AI systems as part of their operations, particularly in campus-type network setups.
The shift towards local AI models is crucial for real-time applications, like robotics in manufacturing, since delays can disrupt operations.
This illustrates the future of physical and embodied AI, which accentuates the need for localized processing to enhance efficiency.
"My role currently at Cisco is the GM for enterprise networking and collaboration."
Anurag, a seasoned engineer by training, transitioned from a software programmer to a leadership role at Cisco, overseeing the enterprise networking and collaboration sectors.
He emphasizes that despite the changing landscape of programming, AI tools now simplify the coding process, easing the burden on engineers who may have been away from writing code for some time.
"AI is disrupting all sorts of workflows and different verticals and industries, and networking is no different."
AI is reshaping how networking works, with an emphasis on software agents and robotics emerging in various sectors including manufacturing.
Cisco aims to leverage AI to simplify network operations and enhance productivity, especially in environments where IT teams struggle with complexity and talent shortages.
"The idea is simple: you can delegate tasks to these agentic systems, and the human role shifts to supervising and guiding."
With the advent of agentic AI, the relationship between humans and machines is evolving from direct commands to a collaborative partnership.
This technology allows IT teams to scale operations by delegating tasks to AI systems while remaining in oversight roles to ensure that operations are executed correctly.
"I think it's important to have a spectrum of control with AI—from having a conversation with it to handing off operations completely."
There exists a broad spectrum of comfort among engineers regarding the level of trust they place in AI for network management.
The recommended practice is to utilize AI for initial tasks and assessments, with humans verifying the results before any final decisions are made. This enhances security and reduces the risks associated with implementing AI-driven changes.
"We have autonomous agents built into our cloud controllers that can automatically detect anomalies in the network."
Cisco has implemented autonomous agents within their cloud controllers capable of baseline monitoring and anomaly detection.
These agents can alert engineers to irregular traffic patterns and recommend corrective actions, but a human review is required to ensure accuracy before executing any changes in the network.
"As these software agents become more capable of performing various tasks, many of us will start using them for our jobs."
The rise of software agents signifies a shift in how work is executed across various sectors, not just in network administration. As these agents assume more human-like duties, their ability to carry out complex tasks will redefine job roles and workflows.
Accessibility to information sources is crucial for these agents. If a software agent is tasked with researching Cisco products, it must have the same access to internal documents that a human worker does, raising vital questions about security and permissions.
The emerging use of these agents introduces complexities in network environments, necessitating a reevaluation of security measures and traffic patterns—especially as AI agents begin to communicate with one another.
"Edge AI is an emerging trend that brings inferencing capabilities closer to where the people are."
Most software agents currently rely on cloud-based models, but the trend is shifting towards smaller, specialized models designed for specific tasks. This advancement allows agents to operate more efficiently on personal devices instead of depending solely on centralized data centers.
With more devices operating on the network simultaneously due to the deployment of local AI agents, network engineers must account for changes in traffic patterns, capacity planning, and security concerns.
New network devices are being designed with AI capabilities in mind, focusing on scalability, high throughput, low latency, and incorporating security directly into network hardware.
"It's not just about replacing a box with another box; it's about talking an end-to-end architecture."
Companies are facing a transformative phase in their network infrastructure as they contemplate the integration of AI. This involves not merely replacing outdated devices but rethinking entire architectures to embrace AI’s potential.
The future of network security will also change; every network device is envisioned as an enforcement point, understanding its role within the broader network context and applying security policies accordingly.
Organizations are increasingly deploying local AI solutions, which are especially useful in sectors like manufacturing where robots operate. By processing information closer to the point of action, latency issues can be minimized, significantly improving operational efficiency.
"We're announcing a brand-new product that brings AI inferencing capabilities in a very easy-to-deploy form factor."
Cisco’s new product, Unified Edge, aims to facilitate the deployment of edge AI by integrating computing, networking, and security into a single chassis. This product allows businesses to manage their networks efficiently, especially in locations with limited IT support, such as retail stores.
The Unified Edge product offers flexibility with the option to include compute modules with GPUs and networking/security modules. This all-in-one approach simplifies remote management and ensures that AI can be utilized locally rather than relying on distant data centers.
Companies are acknowledging the necessity for localized AI applications, particularly for latency-sensitive tasks. The evolution toward edge AI represents a significant shift from traditional networking paradigms.
"It was very hard to run inferencing on the edge."
The introduction of the unified edge has simplified running inferencing on edge devices, which was previously a challenging task.
The unified edge model allows for a seamless continuum from public cloud to enterprise data centers and edge locations, enabling better integration of AI into various services.
"The jobs are fundamentally changing, and some of those jobs are probably going to get eliminated in that specific vertical."
AI is transforming customer service roles rather than completely eliminating them, as employees transition from low-level tasks to more complex problem-solving and relationship-building activities.
Metrics for success in customer service have flipped; the focus is now on whether agents successfully resolve issues and build customer relationships rather than just the speed of calls.
"It is being implemented today. I will say it is early days, but there are certain industries where this is very real."
AI's integration into businesses is already underway, notably in the customer service and software development sectors, where organizations are deploying autonomous AI agents to automate simpler tasks.
The use of AI in software development is becoming widespread, with developers leveraging AI tools, such as Codex, to enhance coding efficiency and automate complex workflows.
"Designers are using AI to generate markups and visuals, even generate JavaScript code."
Designers are now able to use AI to streamline their workflow, moving from creating designs to directly generating code that can be handed off to developers.
AI is also assisting product managers in writing specifications and focusing on customer outcomes rather than just technical documentation.
"Many other companies are seeing the same potential."
IT operations stand to benefit significantly from AI through AgenticOps, which simplifies network operations and reshapes how teams function.
A recommendation for viewers is to initially use AI to enhance what they currently do, potentially boosting productivity by 20-40%, and later aim to reimagine workflows for even greater productivity gains.
"The reality is we're still in the early days of the AI revolution."
Encouragement is given for individuals entering the workforce to embrace continuous learning and develop critical thinking skills.
As AI evolves, new job opportunities that we cannot yet foresee will likely emerge, emphasizing the importance of adaptability.
"I would lean into some of these things around critical thinking."
Individuals should focus on honing critical thinking and reading broadly, not limited to technical documents.
Fostering a learner's mindset is crucial for success in an evolving job market influenced by advances in AI technology.
"Do not wait for your next job to do your best work. Do your best work today."
"To truly differentiate and deliver value, you have to solve some hard problems."
"Think of it as like a CCIE in a box; it's the brains of that model."
"We have a product called AI Defense, which is squarely focused on solving security problems."
"Our security approach is multi-layered, going all the way from ASICs to the application level."
Cisco's security strategy involves a comprehensive, multi-layered approach, which incorporates both hardware and software solutions.
The foundation of this architecture includes ASICs, specifically the Silicon One chip, which provides high programmability and advanced security features.
Key features of their products include tamper-proofing and integrated post-quantum cryptography, ensuring that only authorized Cisco software operates on their devices and preventing unauthorized manipulation.
"We are seeing the emergence of harvest now, decrypt later attacks."
The potential threat posed by quantum computers is significant, as they could eventually break current encryption methods, putting sensitive data at risk.
Attackers are adopting "harvest now, decrypt later" strategies, where they store encrypted traffic with the anticipation that future quantum computing power will allow them to decrypt it.
In response, Cisco is developing and implementing new quantum-safe encryption protocols to safeguard data moving through their networks, ensuring that even if data is harvested, it cannot be decrypted in the future.
"AI is a great learning tool that helps you get educated about your network."
The complexity of modern networking environments makes it difficult for individuals to manage systems efficiently; however, AI can significantly aid in this process.
AI can function as an accelerated learning tool for new network engineers, allowing them to better understand intricate systems and configurations by asking questions and obtaining immediate insights.
This capability extends to IT teams in large organizations, enabling them to inspect and comprehend security policies and device operations more efficiently, enhancing their proficiency and helping manage network complexity.
