Why CX Teams Can’t Afford to Ignore the Network Layer Any Longer

The network was never supposed to be a CX problem. For years, it wasn’t. Telecom engineers leaned on the PSTN’s built-in inefficiency as a quality guarantee, and the broader industry was happy to leave well enough alone.

That era is ending.

With the UK’s PSTN shutdown locked in for January 2027, organizations still routing contact center calls over legacy infrastructure now have a hard deadline. But as Zeus Kerravala, Principal Analyst at ZK Research, makes clear in this conversation, the bigger risk isn’t missing the migration window. It’s treating it as a straight swap.

The problem is that modernizing means running voice traffic across shared, IP-based networks, exactly the kind of environment where LLM-driven bots add another 200-300 milliseconds of latency on top of existing SIP delay. When the bot feels slow, someone has to explain why.

Right now, most CX teams don’t have the tools to answer that question.

Kerravala points to Cisco ThousandEyes as one of the few observability platforms that can trace a voice interaction across local, cloud, and public internet infrastructure simultaneously, helping teams move from reactive firefighting to predictive network management. He also flags edge AI inference as a fast-moving development that could significantly cut latency for in-store and regional deployments within the next year.

The organizational piece, though, may be the hardest part. CX and IT have historically operated in separate lanes. Kerravala’s view is that the companies that get ahead of this will be the ones that stop managing unified networks through siloed teams.

Transcript

SPEAKER_00 0:08

Hello and welcome to CX Today. I'm Reese Fisher, Associate Editor, and today I'm delighted to be joined by Zayas Caravala, the principal analyst at CK Research. Zayas, thanks for joining me today. How are you doing?

SPEAKER_01 0:21

Hello, Brian Reese. Always great to talk to you.

SPEAKER_00 0:25

Yeah, yeah, and uh I'm excited for today. It's uh something me and you kind of talked about around about a week ago. It's a really interesting topic right now, I think. This idea about the almost the resurgence of the network layer in some ways, which kind of feels like it's back on the agenda of CX Teams in a way maybe it hasn't been for quite some time. On that subject, do you think the era of networks kind of being someone else's problem is genuinely over, or is the network layer still being underestimated in some respects?

SPEAKER_01 0:57

You know, I I think in the communications industry um it's diff it's certainly different than it has been, like in data networking where the network has been so critical. With comms, um I do think there was a bit of set it and forget it attitude, and part of that is because so many of us relied on things like the PSTN and ISTN for so long, where it in a lot of ways the quality was built into the inefficiency, right? But if you look at a PSTN connection, the bandwidth capability of that isn't being used anywhere near its capabilities, and so it's an inefficient network that guarantees quality. But you know, certainly in if you think of traditional data networking, you you build networks, you load them up with as much traffic as you can, and then you you worry about the the degradation of quality as you know as time goes on. And so I I think it's this this crossover of data networking and um uh and IP networking that's been happening you know on on LANs and things like that for a long time. Uh I find it interesting on your the public networks, though, it's some combination of IP networks and and PSTN. And if you think about like if I make uh uh you know uh an end-to-end IP call, it's crossing all those networks somewhere. So it's uh but but I think more and more people want to have one common network, and the only way to do that is with the uh traditional IP network.

SPEAKER_00 2:22

Yeah, yeah, it's interesting. You mentioned PSTN. They obviously over here in the UK the the shutdown is is on the horizon now where PT OpenReach are kind of permanently switching off the the traditional copper telephone networks. Um that's coming in January 2027, I think the latest details say. You know, enterprises still rooting contact center calls over PTSN hard, you know, they have a hard deadline now to migrate to SIP. Is this a do you think a genuine opportunity to rethink contact center routing architecture, or uh do you think maybe most companies are still just gonna do maybe a a like for like swap and call it, you know, calling it a day?

SPEAKER_01 3:03

Yeah, it's a good question, right? And so I do think with um uh with a human-to-human call, a delay of about 200 milliseconds is barely noticeable. And our SIP networks and PS networks are built for that. So um the problem becomes the traditional SIP configurations, they were designed for human speech, so you do get that um, you know, the the 200 milliseconds, but then when you start to mix in an LLM, now you're adding about another 200 to 300 seconds of mill uh milliseconds of latency. And so when you combine the two, now all of a sudden it it becomes um it's you know, it's not a huge amount uh of latency, but it's becomes noticeable, right? And and so I think there's a couple of ways to handle this. The first one would be if the PSTN shuts down, you know, what the the the government's doing is they're forcing companies to face their infrastructure debt. And one way to do it is you could replace all your ISDN connections with SIP connections one for one. That shouldn't be the goal. You shouldn't be kind of taking this new technology and make it look like the old technology. It should be how do I modernize my network and do a bunch of different things? And so if we're gonna do that, we're gonna be running on a network that carries other traffic and more traffic and shared traffic and things like that. And so when a customer says the bot is slow, the CX team then has to have the right tools in place to be able to go see why the bot is slow, right? And so I think here is uh this is gonna create a fundamental rethink of these voice networks. And I actually think, Reese, there's a good chance you see voice traffic go up in the era of bots. I know for a long time we've been talking about voice dying, and nobody uses that. But if the I've always said that the only interface that we're all born with is voice, right? And so if it works well, we're going to use that as de facto. But make no mistake, for the telecom engineers out there, this creates a, you know, like I said, a fundamental reset and re-architecture of how they design networks and how they think about them.

SPEAKER_00 5:15

Yeah, yeah. It's interesting around voice. And again, I think we talked about this a little bit last time we chatted, but it I'm definitely seeing it a lot more than desire. They seem to be prioritizing voice again. I know part of that is voice AI, obviously, that's an avenue, but I think even outside of that, it feels like it's more on the agenda than it was certainly a year or two ago from the conversations I'm seeing. And yeah, this will be another another reason that could push it further up the agenda again. So it'll be interesting to see how that plays out.

SPEAKER_01 5:44

Yeah, I I I think it's uh you know, these are skills that a lot of telecom engineers aren't, you know, haven't had to have historically. Like is it the the beauty of the PST and it was inefficient, uh, it was they managed quality through inefficiency, right? And so the you know, data networking doesn't work that way though.

SPEAKER_00 6:03

Lost up an old phrase for it. Yeah. Um I did also want to mention uh Cisco 1000, because I think it is quite interesting within this debate because you know it's it's repositioning network observability as kind of a CX capability. But do you think a CX leader kind of realistically going to own that conversation, or does it still, do you think, probably get stuck in IT ops?

SPEAKER_01 6:28

Yeah, it's a good question. I think it becomes a combination of the two. Um, and you know, certainly if you if you're not familiar with Thousand Eyes in this realm, it's one of the few observability tools that actually not only spans your local network, but across the internet. So they can tell you if you're you know if you're trying to do some cloud computing activity through AWS, where in the world that slowness is, right? And so I think with CX interactions, um, we're not troubleshooting CX by looking at a red-green dashboard anymore. In fact, when you think about the evolution we've had in networking, we build these things in a way that are so resilient now that you rarely have outages, right? It's more the fact that things aren't working properly in those brownout situations where you know things are, you know, as they say, too wrong for too long. And so it's not, you know, and that's where the observability part becomes um you know more important because we're not looking for things that are up down, we're looking for things that aren't quite working right, and we want to know why. So if a customer says, you know, that interaction was very poor because there was too much latency, you need to find out where that is. And so tools like ThousandEyes allows the CX team to see the entire path of that voice journey and then manage the network and understand where and there it needs to be upgraded. And so I think it actually helps with even pre-planning and things like that. But your question, um, I don't know if the CX team is going to wholly own this. I do think they'll have a role in it, and but uh I think it'll be uh the the best organizations will have the operations team and the CX team work together because if we're using these things for customer interactions, right, that has to become our priority, making sure that those qual those calls maintain high quality. And uh, unlike things like messaging and email, where if there's a little bit of latency, nobody notices. With voice, if there's a little bit of latency, everyone notices.

SPEAKER_00 8:27

Yeah. So do you talk about CX and IT having to work a little closer together? I suppose traditionally that hasn't always been the case. Do you think that this news will help bring that closer together? Or do you can do you maybe think that'll continue to be maybe a bit of a stumbling block for companies?

SPEAKER_01 8:45

Well, no, I I think the the the a lot of the leading organizations I've talked to are are trying to make networking part of other teams. You see networking security come together, OTIT come together, and I don't really see why, in this case, CX and IT won't come together. If, like I said, if you look at industry data out there, right, that it's very common now to see um reports that show of you with one or two bad interactions, companies will drop loyalties to a brand. You know, people are spending a lot of money on rolling out these AI systems. And um, and even if you're not using AI, right, you still spend a lot of money managing your contact center and you're gonna go through a whole network upgrade. If the problem now becomes the network, then all that money that you've put into modernization, you're not getting the ROI because the network's not being managed, right? The only way to do that is to bring those teams together. And so I think it um it it does, you know, that's it's it's kind of happened in some other areas, like you think voice over IP in the local network. This is just to me an extension of that, but it's something companies should be wary of and needs to be driven from the CIO down.

SPEAKER_00 9:55

Yeah, yeah, definitely. That makes a lot of sense. Another area, I guess, when I'm reading around this subject I want to pick your brain on a little bit, is I used is edge deployment of AI inference is being talked about as a way of kind of bringing processing closer to the network and cut latency. Do you think that's a realistic option for contact centers right now, or is it still mainly kind of an experimental phase?

SPEAKER_01 10:21

Uh I think it's um it is somewhat experimental today, but as we've seen in the this world of AI, things tend to move pretty fast. And uh, you know, if you follow the news coming out of GTC, uh NVIDIA CEO Jensen Huang declared this the age of inferencing, in which much of that stuff will be done at the edge. I do think um, you know, for CX teams, the goal would be to minimize the amount of latency in every transaction. And the way you do that is you move your processing closer to the user. And so if everything's got to go to the cloud and back, right, that takes a long time. Um, and so not everything needs to be deployed at the edge, though. And so I think you'll see a lot of like in-store transactions, things like that happen at the edge. You'll have regional hubs set up for a lot of maybe regional processing, like a large company could have an AMIA hub, but then you'd use a centralized uh cloud, uh, you know, global cloud for other types of data. And uh um, but I I do think, you know, if you think of the example of I always use the ultimate edge devices, a self-driving car, if I'm issuing voice commands, um, you know, I really want the car to process that data. So if you think about even uh like in retail, they have smart mirrors now, you know, it's stadiums, there's a lot of in-store in stadium tech, right? All that you would want to be processed locally at the edge. And so edge computing's coming, um, it's coming pretty fast. I think it's a lot of experimenting today, but I think if we have this conversation in a year from now, uh Reese, uh it'll it'll be mainstream in the CX world.

SPEAKER_00 11:58

Yeah, yeah, that's a good point. Like you say, with AI, it is things are absolutely moving a hell of a lot quicker than they have in the past. So yeah, I think your own prediction.

SPEAKER_01 12:08

Fastest moving technology I've seen, and I've been and I've been covering tech a long time, so yeah, yeah, exactly.

SPEAKER_00 12:15

I guess just as a final question, Z. I know you you've kind of covered this throughout the discussion, but if you're an organization kind of looking at the network layer situation right now, what what advice would you give them? What steps would you recommend they should take?

SPEAKER_01 12:30

Yeah, the first one is you need to stop thinking about your network as separate networks, right? And often you go to companies and they'll this person manages our WAN, this person manages the LAN, this person manages the cloud network, right? It's one network. Uh, you know, CX transactions don't care that you've got four different WAN, you know, network managers in there. And so you need to think about that transaction from being um, you know, the complete end-to-end. And you can say from the mobile device to the cloud or however you want to think about that. Um, it's one network, and companies need to orient around that. I think second, um don't focus as much on up-down type tools that you know, network monitoring tools that simply tell you the status of certain devices. All networks are built today with resiliency in mind. That I would think that in most cases you could go turn something off and nobody would notice, right? But it's those um performance issues that are harder uh to manage. And I think that's where observability tools become important. And so if you're gonna think about your network as one network and you're gonna bring in observability tools, make sure the observability tools are able to span that entire end-to-end network. And uh third race is you need to make sure that the organizational team that supports it under, you know, is it it has been those teams have been brought together, right? I think the worst thing you can do is try and manage, you know, a unified network through silos of people. And um, you know, I I think that's it's it's very difficult to do that because you end up with a lot of finger pointing. And I guess I can give you one more point. Leverage AI, right? We're leverage AI to help you understand what's going on. There's so much telemetry data being brought into these networks today with different tools. Um, you know, like I said, you brought a thousand ice, but Cisco's got a whole bunch of different observability tools, as does do a lot of their peers. And so people can't ingest that amount of data fast enough to understand what's going on. And if you do that properly, AI can actually help you move from reactive troubleshooting where the user is complaining about you know the transaction being slower to be more predictive. And so, you know, let's say over time your call volume continues to go up. Well, AI can predict at the moment at which your your uh your call quality will degrade to the point where users will complain. Use AI to help figure that out, stay ahead of it, be predictive and proactive versus reactive. That's a much better way to work.

SPEAKER_00 15:10

Yeah, absolutely. I think that's uh that's a great place to leave things. Thanks, Yes. This is a really important topic, but it's also I think a very technical topic. So I think it's always helpful to speak to people like yourself who have deep expertise and it's a very important thing.

SPEAKER_01 15:23

I'm glad you're doing this because I do think this part of the transformation is going to fly under the radar for a lot of companies. Like I said, they'll probably go through the migration of PST into SIP, not fully understand the implications of what they're getting themselves into. And so I think, you know, this it uh, you know, kudos to you for thinking this topic up because I do think it's important when you want to get ahead of this and have this in place before you do your migration.

SPEAKER_00 15:48

Yeah, yeah, absolutely, completely agree. Uh so yeah, just thanks again to us. Really, really great chat, really enjoyed it. Um, I also did just want to quickly thank our audience for tuning in. Uh, remember you can like and subscribe to the channel if you enjoyed this chat, and you can head on over to cxtoday.com for more stories like these. Until next time, thanks for watching.