Clarence Reynolds, TelecomTV (00:07):
The integration of artificial intelligence into telecom is shaping core facets of the network. And here to discuss AI's transformative power are Abhishek Shankar, president of Tech Mahindra, Raj Yavatkar, CTO of Juniper Networks, Hugh Crean, chief Architect at Three Ireland, and Jason Hogg, GM for AI ops within the Azure for Operators Group at Microsoft. Welcome to all of you. So I want to start with you, Hugh, because you are actually at the front lines of this process. How has AI influenced customer service and your operations? Yeah,

Hugh Crean, Three Ireland (00:49):
I mean, the whole area of AI is very exciting, especially for call center. Obviously there's a lot of hype, there's a lot of solutions in the market. Most mobile operators already have some sort of chat solution. Probably the first entry of AI is those chat solutions are now driven by ai. Next generation IVR solutions are, in my opinion, going to be transformative. Many, many of them come with dynamic menus, visual IVR capability and ai virtual assistance. So from a call center perspective, it is going to have a huge impact. It's not even so much having AI agents dealing directly with customers. It's human agents working collaboratively with virtual assistants and automating a lot of the back office tasks that an agent does today with typing up notes, summarizing cases, raising trouble tickets. A lot of that can be automated and the human can focus on, I suppose higher value activity.

(02:00):
So it's an exciting time from an architecture perspective. We are cognizant that we need to move, but move carefully. There's many cases of AI chats going rogue, et cetera. So we want the governance around it. We want control around it. We've all invested a lot of money putting a 360 view of customer in front of the human agent, we need to make sure we keep that for the AI agent. So we don't want different silos of ai. We want to keep it governed, fed with a common data set so we can control it. But yeah, definitely it's going to be an exciting, I'd say within the next year there's going to be a massive change.

Clarence Reynolds, TelecomTV (02:47):
And Raj, how about you? How has AI impacted network operations?

Raj Yavatkar, Juniper Networks (02:52):
So traditionally, if you look at network operations, telecom operators would use multiple dashboards to look at different aspects of the network and try to monitor and troubleshoot networks. That's a very manual labor intensive process. So what we have done is that we took telemetry from all the aspects of the network, starting from end users, network, fabric, network devices. We all collect that in a public cloud and apply machine learning to do correlations across us. So we can automatically find out any animal leads or any troubles. And before those are noticed by the operators, we actually are able to troubleshoot. So one of our customers has publicly stated that 90% of their troubleshooting tickets get self resolved. Nobody has to intervene. So that's the power of artificial intelligence and machine learning that we are beginning to see network operations.

Clarence Reynolds, TelecomTV (03:46):
And Jason, how does your company see AI impacting operations as well?

Jason Hogg, Microsoft (03:53):
It's amazing. There's probably not an area of operations that won't be affected by ai. The story that Raj just described is very similar to the journey that many of our customers are on. The interesting challenge, of course, is that the journey starts with data helping operators break down the silos that have actually existed over the years by virtue of the data coming through different network functions, not necessarily having great adherence to common standards. And the cloud's a great sort of great tool for actually breaking down the barriers. Within Azure, we've released a new offering called Azure Operator Insights, which is all about helping operator scale of data, get ingested, moved into cloud, get managed using a modern data mesh architecture, enable and supervised machine learning techniques like anomaly detections. You can look for sort of idiosyncrasies and sort of patterns that you may not notice otherwise.

(04:42):
But most importantly, again, similar to what Raj was saying, take the tickets that usually existed in operating operations rooms or customer support teams and actually help validate that the actual issue is real. Make sure that the way it's prioritized is real, and then go back and look at previous RCAs troubleshooting content and actually have mitigation guidance integrated into the ticket so that then your operations seems so much more effective and they're able to actually start thinking about how to actually enable better customer experiences or ideally start thinking about sort of new offerings. Absolutely. So it's very similar

Clarence Reynolds, TelecomTV (05:19):
And you have a very unique view from the helm of the ship. How do you see AI impacting the competitive environment?

Abhishek Shankar, Tech Mahindra (05:29):
Sure. So from the helm of the ship, I'll probably throw some statistics to you. What I had an opportunity was to look at several AI proof of concepts, and I just want to differentiate and talk more generative AI and not predictive ai, which both my colleagues just talked about. From a generative AI standpoint, what we are seeing a competitive advantage is really around call centers and customer care. So seven out of 10 experiments today, the way we see globally at across our client base are in the broader customer care contact care space. And about three out of 10 are in developer productivity. But I do think that when we meet next year, at the same time, the conversations will be on generative AI on operations and specifically network operations,

Raj Yavatkar, Juniper Networks (06:13):
If I may. Yeah, you're absolutely right. But generative AI is not necessarily only for call center. We are beginning to see that in network operations, for example, since you get a natural language interface, you can now talk to the network to query the status or give instructions rather than trying to understand the nitty gritty details of scripts, writing scripts, manual operation. So that's beginning to happen. We are already starting to apply that for our products. And like you said, in a year from now, we'll come back and admire the rate of progress is just amazing.

Jason Hogg, Microsoft (06:50):
I think the natural evolution of that will be that the GPT based agents could actually be recommending configuration changes, for example, but you'd be mad to actually deploy them unguarded. And that's why sort of the idea of the copilot sort metaphor with the agent working alongside the AI based agents, maybe making a recommended change the network, but then ensuring that the actual deployment of the configuration change adheres to your safe deployment practices. You've got a phase where to roll it out, test it, stage it, and then deploy en mass. And that's going to be that closing of the loop that we've all been super excited about

Hugh Crean, Three Ireland (07:22):
For the longest time. I think that's big impacts on the OSS layer. So if you think of traditional monitoring architecture where you get alerts flowing down one direction and a human gets on the phone, raises the alarm, there could be an hour before you call people out. So now you can have an AI agent getting that alert. You can have a number of pre-agreed mitigations steps that can be done to mitigate rero traffic, restart process, kill a server if it's in a cluster. So you could have real time intervention, which would actually prevent the outage in the first place. So there's a lot of potential in the OSS side. So

Jason Hogg, Microsoft (08:08):
I think the interesting thing will be to try to, how do we actually start to determine which of those automated tasks or which of those remediations can be done in an automated fashion versus those where you do actually still want human in the loop, so to speak. But

Raj Yavatkar, Juniper Networks (08:21):
There's also learning component as you said, that it'll start acting upon those in a sort of proactive way. But the process to learn continuously, and if you're a network operator and if you have multiple networks, you can learn across multiple networks. So one learning can be applied to other, which is also very powerful

Jason Hogg, Microsoft (08:41):
I think. And the real opportunities are, I think for operators is in general they've sort of operated in the regions that they operate in. And how do we actually take the learnings from one region, share that with operators in another region so that everybody's able to focus the new business opportunities, the new experiences they want to deliver to their customers, I think is the next opportunity. And that's probably something that you are thinking hard about, I believe.

Abhishek Shankar, Tech Mahindra (09:03):
Absolutely. And connecting the tissue Clarence to your question on competitive advantage for operators, I think doing it first versus doing it right, the latter might be more of a competitive advantage than just getting heads on with it. So I think building safety and security and the guard rails around it will be probably one key aspect of competitive advantage.

Hugh Crean, Three Ireland (09:27):
Probably on the competitive side of us, the campaign management is going to be impacted by ai. So today we're, let's say fairly standard campaign, you get an inbound contact, you've got a pre-agreed offer or next best activity that flashes up to the agent. But in AI that can be a real time decision making based on data history, even predicting what the customer could potentially want. So a lot of potential on the campaign side upsell, cross-sell. It's definitely something that we see happening. Also,

Raj Yavatkar, Juniper Networks (10:13):
One application of generative AI in that context is today the sales process is through some mechanics like Salesforce where you put in the information you get on a customer call, you finish the call, and then you remember what to put in. But if you're generative ai, you can monitor the entire call automatically populate the Salesforce fields and also rate the call how successful it was. So you can use it for training again. So sales process is going to drastically change and become a lot more efficient.

Clarence Reynolds, TelecomTV (10:44):
And Raj, you bring up something really important. Have you contemplated what success looks like? Will that change the way you measure success? Because AI is just like Raj has given that scenario, has really changed the way you can look at data. Well,

Abhishek Shankar, Tech Mahindra (11:03):
My broader view on the industry is that it's a relatively simple industry. Success is adding new subscribers and controlling churn. So as long as we can drive a correlation between basic business metrics, which I think we are all in a very good position too, at least from my lens, from the operator side of the world, that is, but from a Juniper lens, you might want to,

Raj Yavatkar, Juniper Networks (11:27):
Yeah, I think one change that could happen that you could measure is that today the operator is the way you operate. Suppose you want a new cable service, you want something done for a B2B enterprise connectivity, you file a ticket, it takes two, three days to get the result. But now with generative ai, with the kind of automation you apply, you should be able to do instant on demand service today. The public cloud operators provide that. I don't to Azure and a pilot ticket and wait to here to get a service where I do that with operators. It's not necessary. Generative AI will change that. So that's a big measure of success in market

Clarence Reynolds, TelecomTV (12:04):
When it comes to the ecosystem that you all operate in. Do you worry that your customers or partners aren't keeping up with AI and aren't going to be ready for this transformation?

Hugh Crean, Three Ireland (12:18):
I think it's the opposite. Every partner has the solution that what they're pushing and want to sell. So it's kind of like actually for us it's about actually deciding what is the best solution or is there a one size fits all? Or back to my point about do we have a single AI engine that's got a 360 view of all the channels or do you actually have multiple, we looked at some IVR suppliers recently and one of them comes with all the AI engines integrated in the back. So they found that some of the AI engines were better for sentiment analysis. Some were better for language, so they actually call different ones depending on the use case. So again, that's the potential there is amazing.

Jason Hogg, Microsoft (13:09):
I think the interesting thing, I think when it gets to the core network, we've got lots of different network functions being delivered by different vendors. Unfortunately there's not great either adoption of or adherence to standards. And what that ends up meaning is that you end up in this segregated networks with siloed data that actually makes it quite hard. The value proposition for creating generalized AI models is just not there because the people that build AI models almost need to build them for each of the different variants of the functions. And I think unfortunately that's one of the things actually limiting the adoption of ai, traditional AI for operators. So one of the things I'm hoping with 5G, and as we start thinking about with six G, a big part of the movement has to be on operators in particular requiring their vendors to publish schemas, publish test data to enable the broader AI ecosystem to actually start to flourish in addition to the obvious GPT based scenarios that were basically starting to see early days for. And it'd be the combination of the data, the different AI models, maybe the generative ai, the multimodel, generative ai, being able to then integrate with other AI systems that have been built and create the unified view that everybody's hoping and hoping for.

Raj Yavatkar, Juniper Networks (14:20):
Going back to your question, there's also fomo, fear of missing out. So people are jumping on the bandwagon very quickly, but some functions like call center are low hanging fruit, in my opinion. Some will be harder to achieve and that's where the competitive advantage will be. The people also start differentiating, but there's a lot more that Jason mentions multimodal model. There's so much development happening, the state is in flux.

Abhishek Shankar, Tech Mahindra (14:48):
FOMO is absolutely, you can see it and it's palpable in every conversation. But I think there's one thing which is different about generative AI versus traditional. I think traditional mist, for example, it has been doing machine learning for years. So some of this has already been there. But when it comes to generative ai, my son who's 10 years old, he tells me more about GPT four and its comparison with others. It's out there in the consumer's hand. So a lot of generative AI is ubiquitously present relative to traditional ai, which we didn't know existed, but it always existed in some way or the other.

Raj Yavatkar, Juniper Networks (15:27):
Yeah, but it's something like MIS ai, which already uses natural language processing, has a head start to start incorporating generative ai because it doesn't take much to do that. That's an advantage and you have to maintain that. But I think to your earlier point, the whole bunch of industries are going to try to transform. I think it's a pace of innovation. We'll decide how quickly we do it.

Jason Hogg, Microsoft (15:50):
I think the really interesting opportunity, I think we've talked about sort of the obvious applications making call centers, more efficient operations teams, more efficient, maybe even sort the dark narc and making operation teams more autonomous. The real question there will be the operators have got this incredible opportunity to actually start integrating AI into their networks at the edges of their networks within the RAN and start offering that to enable consumer devices to use AI that they may not have before. And this is where certain mobile Congress this year was the AI ran alliance was actually announced. And so part of their goal is to start integrating, ran into more parts of the network to actually create new business opportunities, new communication, what would you call it? Opportunities for operators to actually enable

Raj Yavatkar, Juniper Networks (16:33):
And AI for ran the alliance you mentioned is important because RAN is a very expensive resource. People have spent billions of dollars to enable 5G, and they're not being able to monetize it. We all pay the same amount for our iPhone service plan, no matter it's 4G or 5G. So introducing new services that take advantage of AI is imperative. It's really important. And

Jason Hogg, Microsoft (16:56):
The other opportunity is from a sustainability perspective, one option is to better understand where the power is needed and power down the cells. But a better option would be to actually be enabling new scenarios for consumers that you can monetize. So that's where things will get really exciting in the coming years.

Clarence Reynolds, TelecomTV (17:12):
Well, we are definitely at the beginning of something big and we're in good hands with all of you. So thank you very much for being with us today. Very welcome.

Jason Hogg, Microsoft (17:21):
Thank you. Thank you.