Resilient Wireless Networks for AI with Cisco Enterprise Networking
Minse Kim, Cisco’s wireless product manager, emphasized that the AI era is profoundly changing enterprise networking, extending beyond data centers to encompass “physical AI” applications in factories, medical facilities, and dynamic workspaces. He noted that surging demand for AI infrastructure components is also influencing customer buying cycles, with some customers proactively investing in Wi-Fi 7 now. A key insight is that while AI infrastructure is often perceived as data center-centric, the actual consumption and training of AI models, particularly for robotics and autonomous systems, relies heavily on high-performance, low-latency wireless connectivity, making Wi-Fi 6, 6E, and 7 crucial “last mile” technologies. Cisco’s Wi-Fi 7 access points are designed to meet these demands, offering multi-gigabit speeds and backhaul capabilities up to 20 Gbps per AP.
Addressing Wi-Fi’s traditional reliability-versus-speed trade-off, Cisco has developed Ultra-Reliable Wireless Backhaul (URWB) capabilities integrated into its Wi-Fi 7 APs. By dedicating a radio, URWB provides a stable, predictable, and low-latency “wired-like” connection, which is essential for critical applications like robotics that cannot tolerate the blips and jitters common in traditional Wi-Fi during client roaming. Beyond connectivity, Cisco Wi-Fi 7 APs also enhance spatial awareness and location services. Leveraging technologies such as 802.11mc (FTM) and Ultra-Wideband (UWB) with sensor fusion, these APs deliver sub-meter (e.g., one-foot) location accuracy and low latency, resolving long-standing problems in asset tracking and network operations, as demonstrated by real-time asset tracking in an office environment. This ability to accurately digitize the physical world is fundamental for AI analytics.
Furthermore, Cisco is integrating AI into network operations to simplify management and optimize performance. For instance, AI models leverage telemetry data from 35 million Cisco APs globally to intelligently manage firmware upgrades, learning from customer rollback decisions to improve future deployments. AI also enhances Radio Resource Management (RRM) by moving beyond simple rule-based engines to intelligently optimize RF configurations, leveraging historical interference patterns and dynamically adapting to environmental changes to maximize network efficiency and stability. Cisco is even introducing the concept of APs acting as “synthetic clients” to proactively collect network statistics and provide informed recommendations. This comprehensive AI-powered approach, delivering ultra-reliable, high-speed wireless, precise spatial awareness, and intelligent network automation, is not a future vision but a current reality, with thousands of customers already using Cisco’s AI-powered network solutions.
Presented by Minse Kim, Product Manager, Wireless, Cisco. Recorded live at AI Infrastructure Field Day in Santa Clara on January 28th, 2026. Watch the entire presentation at https://techfieldday.com/appearance/cisco-data-center-networking-presents-at-ai-infrastructure-field-day/ or visit https://techfieldday.com/event/aiifd4/ or https://www.cisco.com/ for more information.
Transcript
My name is Za Kim. I'm the, uh, wireless product managers. So I've been doing our wifi a long times, and, uh, now I'm doing mostly at the agent tech ops because every Cisco employee is working on the agent tech ops as a north star.
And then, uh, when it comes to the, uh, enterprise network and wifi network, the more cover a little bit about what does it mean for my enterprise network, my campus network, when we comes and see the AI era. In fact, I had some chat on the back with chatting with my customer, and, uh, he's kind of a, coming back, has a, his old COVID pan. I mean the, the, the, the panics.
What that mean is they now see the, uh, demand soaring all the chip set and component. The pricing is, uh, rising. It's dividend price going up crazy.
But they said, okay, I might have to buy my wifi seven right now because the price of my P might go up anytime. So, uh, we see some interesting changes that AI even is changing, the customer's buying cycles, but, uh, it's not about the, uh, ment it's actually happening across the whole domain. And then, uh, when it comes to wifi and ai, when you find that common denominator, one common denominator is a physical ai.
Because we talk about the ai, we always think about the AI infrastructure being something in the data center, right? Or that's data center, data center connectivity, data center, power consumption. But in, in, in reality how we consume those data, how we train the robot, you're probably very familiar with how you train the robot, right?
Mm-hmm. There's operator with the, via Google doing the stops, and the robot mimic it simultaneously while we sending all the seven 10 video feed to the, to the central cloud, the AI cloud to, uh, create a model and, uh, what interface they were using, what type of network technology that we use on that wifi six, wifi six, E and seven. Mm-hmm.
So wifi become crucial point on the entire consumption cycle of, uh, AI ecosystem from the trading, from the, uh, downloading to the, uh, to the actual, the, the, the consumption. So, uh, we, we are looking for wifi to be a sub millisecond latency with a multi gig performance as a last mile. So we talk about the ultra ethernet with a hundred giga more.
Yep, for sure. Because a single wifi access point can easily handle five gig and six gig. Maybe I can talk about the AP topic later on, but, uh, you can find the most of the enterprise vendor.
Of course, we start with the Cisco that we launched the, uh, high-end wifi 71st, starting with the 20 giga PS of the backhoe speed under the AP that placed on the ceiling, right? And imagine that, uh, this type of hotel that have, uh, 500 AP or 400 ap, what types of a network link that you needed, multi terabyte, multi tera PS that's kind of a common expectation that we want to see from the AI era. So, uh, these changes are driving adoption or lot, for example, like a car manufacturing.
All the top five US car manufacturers are using Cisco, uh, including some foreign company that who built a factory in US also using Cisco. I'm talking with them daily basis. And their top interest is, uh, now they environment changing.
Now we have to find a way to work along with the robot and how robot can operate it. Wifi, there's other technology because there, there's no wire on the robot. Similarly, the, uh, the medical spaces, they are the first adopter of the ai, all the images, scanning and readings and the remote services.
They're using AI to train the images and train the model and the where to start from it. Yeah, they have to scan the images with, uh, with, uh, the mobile devices. And then, uh, workspace also being involved because of the, uh, post COVID era, we all back to office.
And then, uh, spatial knowledges and the real estate investment is extremely crucial point of the cost. I mean, cost control. Having said that, are we ready for AI era and agent ops with the wifi seven?
I think that as an infrastructure, the first requirement is, hey, we have to support from the low, low end all the way to the up end. So when you look at the, uh, APS capacity, normally we don't really go by, Hey, is this really good ap, bad ap? I mean, uh, it is a bit difficult to kind of, uh, draw the line, Hey, this is AP that considers the best ap.
So, uh, one of the, if I give it the one rules of thumb, easy way to identify the good and better and best is go by special stream. Special stream is the same as, uh, you know, gas cast cylinder, four cylinder, six cylinder, a cylinder, 10 cylinder free retail engine, free 12 engine, right? So our latest source is V 16.
I don't know any with the V 16, uh, maybe, uh, maybe some big trucks, right? So, uh, 16 special stream is only came from Cisco. There's no other vendor building 16 special stream ap because it's so, I mean, massively powerful, many vendors saying that, yeah, this is too powerful for my, my taste, but the Cisco building, yeah, can you do that?
Oh, oh, sorry. Sorry. I can, I can, sorry, I cannot move it.
Yeah, I have to stay here. So, uh, 16 special stream all the way to through special stream, it fits everywhere. But the point is this, when you talk about the wifi, wifi is no longer about just, uh, traditional a OTA 11, a OTA 11 we've been using for the last 20 years, and is we really designed for zero latency, zero roles, 10 gig, 20 gig throughput, and data that AI infrastructure needed May or may not, right?
Because the wifi has been what is the biggest problem with wifi is a speed or reliability. Reliability. So most of the time my wifi have to post a hundred Mac.
I don't really have much of a remorse. Well, I have, I have a question about that. Yes.
I'm noticing kind of the, the lack of mention of the ultra reliable wireless back haul. Yes. Yeah.
Yes. That's You're Talking about reliability. Yes.
That's what I'm coming, that's what I'm coming, yeah, that's the next slide that I have it. So, uh, what we found was the, uh, wifi protocol itself has to be evolved to provide a more reliable connection. But we can't wait for industry because now we realize that industry talk starts talking about it in wifi eight and wifi nine.
Alright? I mean, we are looking at, we are talking about the AI right now. Do we have to wait another four years, another five years, another 10 years for client to be ready?
So let's get it ready. Let's deliver the outcome to them user. That's why that all, all our Cisco wifi seven AP capable to operate ultra reliable wireless backup.
And I'll show you the demo that how does it changes the entire landscape of a wireless experiences right next to it, right? When talk about the wifi, wifi is always unpredictable because, uh, wifi connection is a stateless, there's no connection per se in terms of wifi. So that's why that is always unpredictable.
Sometimes connect sometimes is connect because there's a no, um, stringent connection control. Uh, and then, uh, even the media access, random rather right? Is very opportunistic.
Mm-hmm. So, uh, in the ether and world, that was another issue. But the ethernet, they increase the clock speed.
They did, they have a dedicated medium switching. So it's fine, but in wallet, it is shared medium. In the shared medium.
Without stringent access control, you can't really have a predictable connection, which is bad for robotics, which is bad for any other autonomous system. So what's the best way to handle it, building the next generation wireless? I will show you the demo that how does it difference between two?
So, uh, essentially we have a two AP learning with a wifi as well as ultra level wireless back home. So because of the Cisco AP have so many radio, I told you that we have 16 special stream with 16 trucks, right? So, uh, we dedicate the one radio into wifi, other radio into ultra level wifi, I mean wireless backup.
So, uh, and then, uh, let's imagine that, uh, how your new physical AI board operate in the, uh, factory spaces or workspaces, uh, along with you. So, uh, we place the, uh, the little wheel, oh, sorry, wheel on computer. Yep.
We have a typical general AP on the ceiling, right? And then, uh, my TME, the Chris, I was, uh, kind of learning through with that little, uh, I mean the, the launch table, uh, then catering. So we have a two system, one system on the, uh, uh, right hand side showing wifi statistics.
On the other hand, ultra level statistics on the left hand side. Look at all this jitter, look at all these jitters and, uh, the fluctuation between two. You've noticed a big blip on the wifi.
Mm-hmm. Why that happen? Yep.
Because the wifi only can cover so much of area if I, I mean, put like a, I mean 30 feet or 20 feet away from the AP client need to join the new ap. So every time that the client get disconnect and reconnect, there is a blip. And the blip might be as short as a 30 millisecond or as long as a two three second of a delay.
And that can be really, uh, I mean the, I mean the life and death matter, right? So, uh, but when you look at the, uh, ultra level wireless backhoe, it's like a switching ethernet. It is like a freaking one gig switching port, right?
You don't see any blip, you don't see any, um, jitters on the latency so that you can almost as if we use your board physical AI devices as a connected devices and wire the experiences. So this is what we are looking for wifi in next gen. And the great news about the Cisco is, uh, you don't need to wait for next another next gen because Cisco is ready to provide it.
And then, uh, many of our factory spaces are expecting this to be part of their next gen infrastructure, and they are working with, uh, key manufacturer, glad that the US a lot of coming back. And then, uh, we are, we are talking with, uh, pharmaceutical and, uh, logistics, and they are, we are expecting a huge boost of the, uh, new way of connectivity control in the, uh, AI infrastructure. What's next?
Wifi is a great, but, uh, what we built is the wifi infrastructure as a multi force wireless connectivity solution. I don't know if you are keeping up with all these GA uses and, uh, uh, apple announced the new Apple L tech tool, right? So while which they provide, uh, the longer latency and longer coverages and better accuracy, longer battery time and all that, right?
And I was thinking that, okay, I mean, the consumer industries are evolving that, uh, they're making the UWB as a part of the day-to-day. Uh, I mean the, the past, why don't you use the u WB in my enterprise network? So, uh, can you make usage of U wb, uh, using my existing wifi infrastructure or wifi plus UWB and says yes.
So this is another, uh, great leap from the traditional location services, spatial awareness, because in wifi world, the location accuracy always been a primary blocker that give us, uh, I mean the bit of remote that the, sometime the location accuracy may be, okay, it might be five meter or seven meters away from the actual location maybe, uh, I mean 10 feet away from the actual location and checking for performance is also not great as it's supposed to be. So, uh, we have to solve the two problem fundamental problem on the location services. One is the location accuracy, second one is location latency.
So how we can solve all that with, uh, proper pricing and, uh, infrastructure support so that all of our 20 or almost three decade old problem resolved with a Cisco wifi seven ap. Here's the one example that we use when you manage it as a NetOps, right? The biggest difference between the, uh, wifi versus switching and backend on the data center, most of the NetOps team do not really know where is my AP installed?
Why? Because AP is installed in the 5,000 miles away by third party contractor or subcontractor, right? Mm-hmm.
We never know, really. We only give them a little blueprint from the Echo and haina, okay? Place the AP right here, and then, uh, someone, they, they take a picture, but they can't send a picture of a hundred, right?
So, uh, location has been always a bit of a deterministic based on the certain assumption that yeah, probably the user my in install somewhere there. And what we have done was we are making use of, uh, the latest and greatest technology of location, uh, ATA 11 mc or name FTM find time measurement or UWV. And we actually do the sense of fusion to really create the, uh, I mean the, the location detection in the, uh, sub meter like, uh, one feet.
So, uh, this was another the great great initiative. I mean, the innovation, every NetOps wifi professional was looking for this feature for the last two decades. So I'm one of them.
I'm so glad that now we are able to pull that, uh, trigger and then it goes beyond just the net talk to net and AP two ap, what, what about the asset, right? So let's look at the other, the demo. So how the asset tracking is also evolved in a new wifi seven era.
So, uh, this is our SF 12 office, uh, San Francisco office where my desk located. And then, uh, my friends, uh, he's testing on the, uh, he's looking at the, some of these assets he left off. So, uh, he, he open up his browser.
I mean, he doesn't need to install anything. He just open up the browsers and I mean click the QR code. He got the location of the current whereabout and, uh, find out, okay, I, I left my, uh, asset on the, uh, on the, the, the north corner of the building.
So it's walking through, the water is walking, there's real time tracking is happening. So, uh, you might have seen this demo a few times, but uh, here is, uh, the true innovation is, uh, coming on the later part of the, I mean demo, just look at it, right? Step by step is tracking you.
I mean, in real time, it's, uh, and then I mean, we observe it, the whereabout of the users, but at the same time, the user is actually tracking and searching for his missing asset, right? He's not looking for the meeting room, which is a fixed location. He's looking for the asset that he just need to find out where he left.
So he is continually walking down the, uh, alley and then, uh, when he reached the, uh, the close the alley, let's see what we, what he found. Alright, there's a notification a lot. Okay, you, now you, uh, you must be nearby the, uh, the tag, right?
So he found it and that this is the true innovation. Let's say that, let's move the tag slightly like it is like one feet next to the original location. You see that the, the a location tag location was a switch immediately.
So this is the innovation. We have this team similar solution from maybe air tech with a one consumer guy with a one bag luggages. But now imagine that you can do it in the scale of, uh, thousands and tens of thousands and millions of assets.
So whole world of nets operation will be changings to the new world. And, uh, imagine the, what types of, uh, spatial awareness and asset analogy, uh, translate into the AI world, right? When it come to the, uh, the AI analytics, first thing that I have to have is a data set, right?
We need to have a data, we need to have a, the physical world model that digitize all my asset and who can help and what can help wifi seven infrastructure, it understand the existing whereabout of the humans asset and the system. And then they're able to, uh, populate it on the digital world. And the set is using the, the AP as an omni link and then the Omni can be done by wifi, BLE, ZigBee dread, UWB, you name it.
So many I was so versatile. So, uh, this is how we changed the whole thing. So lemme talk about the third topic.
I dunno how much time. So I'll talk about the third topic, which is, uh, what mean by AI from the NetOps point of view, the net NetOps point of view, they have a couple things. The first one is, uh, they have to keep up to date changing management.
And you know, what is the most common change in management in the natives world formula upgrade? Because even though that someone never changed the VLAN for their entire lifespan of the certain branch, they probably have to upgrade their phone time to time due to our certain security vulnerability due to our certain new feature that I'm looking for new certain policy changes. So we have to fix the formula.
One thing that we have done was we build the AI model inside of the system how we can, how is it possible? Because the Cisco have 35 million AP on the customer side and 35 million AP with the various generation and AP time, we are collecting the real time telemetry. We are collecting that vendor.
AP got opt down, I mean the crash it and why crash it. And we use all this data to push imagery in intelligently and smartly. And we also make use of the customer's rollback decision.
So when the customer get the rollback, it flags a separate tagging and uh, we select, we engage the people to uh, look at the why the customer rollback the code and so on. So this whole ecosystem done in the orchestration of the, uh, the AI operation of the DevOps and the CICD, our continuous integration and continuous development. Other one is, uh, the RRM, right?
When you talk about the RF and wifi, wifi config never stand in static setting, wifi configuration always changing by every minute, every 30 minute, every hour. Why? Because RF is keep accommodating the new changes.
And then now what, how AI changed that whole landscape. Previously channel change was done with a very simple rule engine. So it keeps, forget what happened the five minutes ago.
So it keep coming back to all the interfer location, which is bad using AI in market, it avoid it. So it intelligently optimize it. So final outcome, you have a minimum change with a maximum outcome, which is all possible thanks to the ai.
The last one, the what if, if we convert AP as a board, now we are living in the world of a board, right? Interested the board. So, uh, this is the same story.
Let's make the AP as a synthetic client. So AP was voluntarily, proactively convert itself as a client and associates the nearby ap, collect the statistics and then, uh, allow us to make the, uh, informed decision and proactive recommendation and what to do in your network. So overall, and uh, you can see the entire integration done at the thousand nine dashboard, starting with a client which is AP as a synthetic client, which never done, never happened in the past.
So all and uh, this is the first time in the industry as well, by the way, as a one, that we have a fewer AP learning at a synthetic client connect with the wireless. No one in the industry has done it before. So last was a takeaway for the three.
We talks about the new infrastructure, how the new infrastructure changed the entire landscape, unreliable, unpredictable, slow speed. Wifi is all bygone. We are talking about the, I mean even beyond the multi gig, right?
We are talking about the 10 G access at the AP level with the zero latency, zero loss traffic, including rooming environment. And now we got the, uh, spatial awarenesses and now all the, all the spaces becoming your assets thanks to the, I mean the new connectivity and thanks to new AI infrastructure. And then, uh, last as a NetOps, they also changing, we also support the full AI system of course, but AI system is a final interface, mostly happen at the pool basis, right?
But what if we led AI to optimize network all the time And uh, this is not our future. This is actually today we have more than 8,000 ai RM customer using, optimizing 1 million AP across 5 million client. So, uh, there's a extreme, I mean, successor stories around the AI powered nets.
So we're hoping to have a continually, uh, checkout what Cisco is doing in our campus spaces. Alright, I will, uh, hand off to Kira. Thanks.
Excellent. Uh, thank you Minze and thank you to Rahul, Shai and Kenny for some hopefully great, uh, engaging presentations. So just to close out, you know, hopefully you got a flavor for how we at Cisco are using AI to make it better for IT admins to operate and manage networks.
Also, hopefully you got a flavor for how we are building networking so that it can deliver the performance and security that's required in this AI era. Uh, I'll just remind you of the three pillars around the three guiding principles that we're really using as we build out our networking infrastructure today, right? So how are we simplifying operations and using agentic ops to do that?
How we're fusing security into our devices as we build them, not just at the hardware level, but also at the software level. And then how are we making our devices scalable so that for the next era of disruption, Cisco devices will still remain valid? Lastly, there is a AI summit next week happening, uh, here in actually San Francisco.
It's free to register, it's online, uh, it's delivered virtually, uh, on the 3rd of February. And we've got some real powerhouses that will be joining us live at that event. So I encourage you guys, if you have the time to sign up, it's free to sign up and join and, uh, enjoy the event.
Thank you very much.