![\"Retail](\"https:\/\/dmsretail.com\/RetailNews\/wp-content\/uploads\/2022\/05\/RETAIL-ONLINE-TRAINING-728-X-90.png\")
<\/a><\/p>\n<\/p>\n
Let\u2019s be honest\u2014we\u2019ve all been talking about AI like it\u2019s something off in the distance. A thing we\u2019ll prepare for \u201csomeday.\u201d But that moment? It\u2019s already passed.<\/p>\n
AI isn\u2019t coming. It\u2019s here. Right now.<\/p>\n
Not tucked away in a lab. Not limited to data scientists. But embedded into how people are actually getting work done across every function, every department, every industry.<\/p>\n
And here\u2019s the kicker: the network most enterprises are running today? It wasn\u2019t built for this. Not even close.<\/p>\n
The shift that already happened<\/h2>\n
Everywhere I look, I see AI going mainstream. Not in some flashy, sci-fi kind of way, but in practical, everyday workflows. Sales teams using copilots to write proposals. Marketers generating entire campaigns with a prompt. Security teams leaning on models to detect threats faster than any human could.<\/p>\n
It\u2019s not hype. It\u2019s not someday. It\u2019s here.<\/p>\n
AI is fundamentally altering how networks operate, driving new traffic patterns, latency demands, and workload behaviors.<\/p>\n
Let me walk you through a few examples that really bring this to life.<\/p>\n
LLMs in the cloud and SLMs at the edge to get work done<\/h2>\n
Right now, people across your company are using AI-powered tools to summarize emails, analyze data, write reports, and even build code. Most of those tools rely on large language models (LLMs) in the cloud, which means their devices are constantly talking back and forth with AI application programming interfaces (APIs).<\/p>\n
But there\u2019s a shift underway. As GPU costs drop and infrastructure at the edge improves, we\u2019re starting to see a move toward small language models (SLMs) running right in the branch or campus\u2014closer to the user, where the data is.<\/p>\n
Why does this matter for the network? \u00a0Because suddenly there are two completely different traffic patterns to deal with:<\/p>\n
- \n
- Heavy north-south traffic<\/strong> from the user to the cloud<\/li>\n
- New east-west traffic flows<\/strong> from local AI workloads spinning up on the edge<\/li>\n<\/ul>\n
Both types of traffic need low latency and high performance to feel seamless to users.<\/p>\n
Real-time video analytics<\/h2>\n
Cameras are no longer just recording video. They\u2019re analyzing it in real time. Whether it\u2019s detecting intruders, analyzing foot traffic, or tracking safety compliance, the AI-based analysis is happening live, often at the edge.<\/p>\n
That process of recording and analyzing generates a ton of high-bandwidth traffic, and it can\u2019t all be shipped to the cloud. To avoid security risks and allow for rapid response, data needs to be processed locally with speed and precision. Traditional networks just weren\u2019t designed for that kind of load at the access layer.<\/p>\n
Live transcription and translation during meetings<\/h2>\n
Now, let\u2019s say you\u2019re on a Webex call and the meeting is being transcribed and translated in real time. Seems simple, right? But behind the scenes, your device is sending and receiving a constant stream of audio and data to an AI engine. Just one lagging translation engine can throw a meeting off course: misinterpreting key phrases, dropping sentences, and leaving participants confused. Responses slow down, clarity fades, and the conversation quickly loses momentum.<\/p>\n
All of which means high-performance networking suddenly becomes the difference between a smooth user experience and a rough one.<\/p>\n
AI agents spawning AI agents<\/h2>\n
This one blows my mind a little. We\u2019ve entered the era of \u201cagentic AI,\u201d where a single AI agent doesn\u2019t just do something for you by itself\u2014it can recruit other agents to help.<\/p>\n
Ask your AI assistant to \u201claunch a marketing campaign.\u201d It spins up one agent to pull CRM data to create a strategy, another to write copy, another to generate visuals, and yet another to launch and track the content across channels.<\/p>\n
Each of these agents is communicating with systems, data sources, and each other at high velocity and often without human involvement. And the related network traffic patterns? They\u2019re completely different from anything a traditional network was built to handle. We\u2019re no longer optimizing for people clicking around. We\u2019re optimizing for AI talking to AI.<\/p>\n
AI-augmented access control<\/h2>\n
Authentication is no longer just about passwords. Many organizations are turning to AI-powered facial recognition, voice biometrics, and behavioral analysis. These technologies can offer faster, more secure access, but only if your network can support them.<\/p>\n
Say an employee is working remotely and trying to log in to a secure dashboard using voice biometrics. They speak the passphrase, but the system hesitates\u2014just long enough to time out. They try again. Still no luck. Now they\u2019re locked out, the clock\u2019s ticking on a deadline, and they\u2019re stuck waiting for IT to sort it out.<\/p>\n
All of this because the network couldn\u2019t prioritize the traffic fast enough for real-time authentication. It\u2019s a small delay with a big impact\u2014and a clear sign that the infrastructure might not be ready for AI-driven security.<\/p>\n
What this means for the network<\/h2>\n
If you\u2019re thinking, \u201cWow, none of this sounds like what we built our network for,\u201d you\u2019re not wrong.<\/p>\n
AI is rewriting the rules and bringing with it:<\/p>\n
- \n
- Unpredictable traffic patterns<\/li>\n
- Huge spikes in east-west and north-south flows<\/li>\n
- Tight latency requirements<\/li>\n
- An explosion of short-lived, dynamic workloads<\/li>\n
- Agent-to-agent traffic that most IT tools aren\u2019t even tracking yet<\/li>\n<\/ul>\n
The bottom line? The old model\u2014where networks were designed for people using apps\u2014doesn\u2019t cut it anymore.<\/p>\n
Now, your network has to support machines talking to machines\u2014at machine speed.<\/p>\n
If AI has already changed your network, what are you doing to keep up? <\/strong><\/h2>\n
This isn\u2019t about preparing for AI. That window\u2019s closed.<\/p>\n
This is about realizing that AI is already changing the behavior of your infrastructure. Whether your architecture evolves to support that reality will determine if you keep pace or fall behind.<\/p>\n
Because AI isn\u2019t coming.<\/p>\n
It\u2019s already here.<\/h2>\n
To keep up, your operations need more than old-school methods. You need smart, autonomous agents that make everything run smoother and faster. It\u2019s not just an upgrade\u2014it\u2019s how you stay ahead.<\/p>\n
\n
Curious how Cisco is leading the shift to AgenticOps?<\/strong><\/h2>\n<\/blockquote>\n
\u00a0<\/p>\n
Share:<\/p>\n
\n \t<\/div>\n
- New east-west traffic flows<\/strong> from local AI workloads spinning up on the edge<\/li>\n<\/ul>\n