Evolution of network automation

 

 

To embed our video on your website copy and paste the code below:

<iframe src="https://www.youtube.com/embed/rJMRfQIs38s?modestbranding=1&rel=0" width="970" height="546" frameborder="0" scrolling="auto" allowfullscreen></iframe>
Ray Le Maistre, TelecomTV (00:25):
Into our second session — it's the evolution of network automation. Zero-touch operations has been a rallying cry, but the reality has been falling short of the vision. Telcos aspire to autonomous networks that self-optimise and self-heal, yet most remain stuck at level two automation, as measured by the TM Forum. The gap between autonomous network ambitions and operational reality threatens transformation business cases. So how do operators bridge this divide in a practical way? That's what we're going to discuss for the next 45 minutes. Let's meet our panellists — and again, I'm going to ask them to introduce themselves, starting at the far end. Beth.

Beth Cohen, Luth Computer (01:17):
Oh, hi. So I'm a long-time telco person — worked for Verizon for many, many, many, too many years, but also did a bunch of startups as well. So I have kind of both sides of the telco and the startup mentality.

Ray Le Maistre, TelecomTV (01:34):
I'm sure a lot of people recognise you from our programming over the years as well, Beth. So welcome to Windsor. Colin.

Colin Evans, HPE Networking (01:41):
Colin Evans, currently with HPE Networking, which recently acquired Juniper Networks. Previously at Cisco, Alcatel, Nokia, to name but a few. Been around the telecoms industry for 37 years.

Ray Le Maistre, TelecomTV (01:59):
Okay, excellent. Welcome. Anita.

Anita Döhler, NGMN Alliance (02:01):
Anita Döhler, CEO of the NGMN Alliance. We are an operator-driven global alliance, and I also used to work for operators, vendors, and consulting companies in the past.

Ray Le Maistre, TelecomTV (02:13):
Okay. George.

George Glass, TM Forum (02:14):
Hi. George Glass, the CTO of the TM Forum. I've been in that role for eight years now, and prior to that I worked at BT as the Chief Systems Architect.

Ray Le Maistre, TelecomTV (02:24):
Okay. Welcome. And Francesca.

Francesca Serravalle, VodafoneThree (02:27):
Francesca Serravalle, VodafoneThree, and Head of Network Automation, part of the Business Strategy and Transformation team.

Ray Le Maistre, TelecomTV (02:35):
Okay. Welcome back. So Francesca is our DSP co-host for the session. Before we start the panel discussion, I'm going to ask Francesca to go to the podium to deliver a few opening words. Francesca.

Francesca Serravalle, VodafoneThree (02:53):
Thank you. Good morning, everyone. I will start with a simple question. Is autonomous network possible? Yes, it is. We know what we need to do. We need to automate detection, decision, and action, and we know what is required. We need to go from machine learning to gen AI to agentic AI. The next question is: can we as operators realise the vision of autonomous networks? I would say it depends. If we treat it as a technology evolution programme, we are set for failure. If we treat it as a multi-year business transformation programme that spans process, technology, and people, we are set for success. What we need to do is to have an approach that is business-led and value-led. It means that we need to start with a clear map of our pain points across the network lifecycle and a clear view of opportunity across the service lifecycle. And as we know our pain points and our opportunities, the next thing is to build up a use case roadmap and prioritise based on a value-risk framework, so that we have clear governance on how to scale.

(04:15):
There are a few enablers that are important for us to invest in. Before we become AI-driven, we need to become data-driven. And this, I think, was the point of the previous panel. We have great data, but we need to make it high-quality — integrity is key — accessible, and secure. So a data lake as well as a data platform that leverages data management to solve those problems is really key. The other challenge that we have is adoption. Change management is key again. I will touch on a few points. It's about transforming the process — this is not just process mapping, it's transforming the process and driving adoption as well. It's about creating synergy between AI and humans in the loop. Thank you, everyone.

Ray Le Maistre, TelecomTV (05:12):
Thank you. Fantastic. Okay. So thank you for that introduction. We've already heard a reference this morning to levels of automation within network operators, and that might be a good place to start because we hear a lot about this. Where do networks sit today on the TM Forum's zero-to-five autonomy scale? Where do we need to be, and where do we want to be in a couple of years' time? And I feel it's only fair, George, to start with you — from the TM Forum, you must be constantly looking at what operators are doing and where they are on this journey, to use that word.

George Glass, TM Forum (06:04):
Yes. The work that we've done with autonomous networks started out by defining the automation levels from level zero through to level five — from manual operations to full autonomy. We also developed an AN level evaluation tool to allow our members to go in and assess themselves in terms of levels of automation and maturity. You don't do it across the whole network — you do it for particular scenarios. So you subdivide your network into a set of scenarios which represent business processes that allow you to manage your network, and you measure the levels. And typically, what we're seeing whenever an operator does their first sets of AN level evaluations, they get a score somewhere in the region of 1.5 to 1.8. And most of the operators we talk to have an ambition to achieve AN level four.

(07:00):
So it's a full level of autonomy that has AI — it has our concepts such as intent-based operations, closed-loop controls, and other capabilities designed and built into the network from the outset. But they're not there yet. And to transform your network, it's a bit like eating an elephant — where do I start? Which is why we came up with the concept of scenarios, subdividing the network into a set of domains, subdividing the domains by technology, so you can go after each area point by point. Those individual transformational journeys lead to an overall transformational outcome. But typically, as I said from the outset, when a member measures their initial automation levels, they're scoring in the high ones or at best low twos to start off with.

Ray Le Maistre, TelecomTV (07:52):
Okay. So that's interesting because especially in the last few months I've heard a lot of talk about level four — but we'll come to that. Now, Beth, this is a topic you've been looking at for years, long before the AN levels. But it's always good to have a measurement. Where do you think the telco sector is in terms of actually achieving measurable levels of automated network operations that actually make an impact?

Beth Cohen, Luth Computer (08:28):
I think it is an important distinction to understand that machine learning and AI are actually not 100% in alignment. A lot of people kind of get them mixed up. And I think telcos have been working on the automation side for many years — that was mentioned in the previous panel. So many telcos are struggling with the concept of: well, okay, it's all about automation — where does this AI stuff fit in? They understand machine learning — that's pretty straightforward because it's an extension of automation. But AI is predictive, as was mentioned earlier, and that's a hard concept for many telcos to really wrap their heads around.

Ray Le Maistre, TelecomTV (09:20):
And do you think that's maybe halted some of the progress? There's a lot of rethinking going on, do you think?

Beth Cohen, Luth Computer (09:26):
I think there is a lot of rethinking going on. And of course, again as was mentioned in the previous panel, there's the proof of concepts — where you always have your best data scientists and network engineers working on them. And then when you roll it out to real life, those aren't necessarily the people required to make this work day-to-day. So I think it's important to understand that part of what needs to be built into these solutions is a high level of automation so that it can work for the junior network engineers who mostly populate the telco's workforce.

Ray Le Maistre, TelecomTV (10:12):
Okay. Colin, you must be seeing quite a lot of efforts in terms of network automation at the companies you work with. Where do you think we are in terms of the progress here?

Colin Evans, HPE Networking (10:26):
Well, it's an interesting discussion. Networks are absolutely becoming incredibly complex. And let's not forget that when we're talking about multi-domain, we're not just talking about the network — we're talking about the compute layer, the virtualisation layer, the applications that sit on top. All of those things need to be coordinated. You're talking about networks that potentially have hundreds of thousands of physical and virtual connections, and on top of that, potentially millions of flows to manage. So that complexity is without doubt something that is almost beyond the human ability to understand and visualise. We're absolutely seeing people using automation to address those kinds of issues. The other thing is: humans make mistakes. Humans get tired. Humans are fallible. They get distracted and they make errors. Most network outages are caused by human error — people misconfiguring things. Tools don't make those mistakes.

(11:30):
Of course, tools have to be designed and built correctly, but they don't make the same mistakes that humans make. So I think we're seeing a huge amount of work on the use of automation to try and alleviate some of those things.

Ray Le Maistre, TelecomTV (11:43):
And are you seeing operators set goals and talk about levels of automation that they're at in particular domains?

Colin Evans, HPE Networking (11:51):
I think each operator will be completely different based upon where they are, and there's a huge amount of organisational change that needs to happen as a result of these automation journeys. Depending on the maturity of the organisation, that determines where they'll be on the journey. So yes, they're all trying to move from — as George mentioned — level 1-something to 2-something. But whether they're ready to move to level four is another issue. That's maybe a few years away yet.

Ray Le Maistre, TelecomTV (12:17):
Okay. Francesca, going to put you on the spot here. Are you beyond 1.8?

Francesca Serravalle, VodafoneThree (12:23):
We haven't done the formal assessment. If I had to give my opinion, I would say we are around two and we are incubating capability for achieving level three, which will probably be within a few years. I believe a lot has been done using traditional machine learning and time series analysis for automating detection. For example, we are leveraging machine learning where we correlate market data, network data, and consumer data to drive smart investment decisions. We're using machine learning a lot to drive AI-driven energy management — we have an energy platform where we ingest meter data, traffic data, config data, and building data, and we automate the energy forecast, energy optimisation, and building validation as well. So energy is probably the area where we are quite ahead in terms of innovation. And of course most operators are using machine learning to do churn analytics, identify likely churners, and run some targeted campaigns.

(13:42):
We are also incubating gen AI into our business processes — especially financial processes — to leverage unstructured data like POs, invoices, and contracts and realise a vision of zero disputes. We're using gen AI to automate HLD and LLD, create blueprint architectures, and we treat it as a transformation programme. As I was saying before, it's not just a POC proving that the technology works — it's more like an MVP. We make a hypothesis on the business outcome and then make sure it's achieved, so that we know we can build a business case for scaling. So I would say we are preparing for level three, incubating capability around gen AI, data lake, data platform, and analytics platform — those enablers that require investment. As DT was saying, it's the use cases that deliver the business outcome and the ROI. But it's still important that we invest in the key enablers that are critical for both level three and level four.

Ray Le Maistre, TelecomTV (14:52):
Okay. And Anita, the NGMN is an operator industry body and automation impacts pretty much all of the work you do. What kind of conversations do you have at the NGMN? What are your members talking about in terms of the importance of automated networks and where they are in that process?

Anita Döhler, NGMN Alliance (15:24):
So we are an operator-driven alliance. We also have vendors and academic institutes in the NGMN, and for the operators we already started in Autumn 2023 with the Cloud Native Manifesto. For the operators, an important part is — as a precondition for autonomous networks — achieving cloud native maturity and cloud native interoperable networks. In addition, what we discussed and also published just recently is a framework for operators on how to move to agentic AI-based operating models. And this is not in competition with the TM Forum's work — we cooperate, to be very clear — so it complements it. In this work, the team used as a baseline the CNCF's cloud native maturity models and added dimensions like people skills, process governance, and technology, of course, providing a framework with five different levels starting from very basic automation — simple use case examples like AI-supported documentation — all the way to truly autonomous scalable networks.

(16:53):
And I think this reflects the fact that our members — and everything we do at the NGMN, as for most industry alliances, is membership-driven — so if our members put effort into this work, it means it's important to them. And one important element is always to ensure that the impact is measured — to be clear about what the business objective is. As Francesca's speech related to: what is the business objective? Is it energy efficiency? Is it network efficiency? Is it to improve MTTR? All those different aspects. And yeah, I think it's a complex subject. I also liked a lot what the first panel outlined — and that it's hard to measure the impact of technology enablers — but it's important to find a way to do that.

Ray Le Maistre, TelecomTV (17:57):
Yeah. It's also important to identify all of the building blocks that enable these processes to happen. And Beth, you've been working a long time — at Verizon, but also with different industry bodies on the development of all sorts of things, going way back to NFV orchestration. Which are the building blocks required right now at telecom operators to enable them to even think about getting to a point where they can have almost zero-touch automated networks? A lot of people are talking about level four, where there is still some human involvement. Nobody's talking about level five — the completely dark network — but level four, hearing a lot more about plans, trials, certain things going on, including in Europe. What's actually needed to be able to put these things in place?

Beth Cohen, Luth Computer (19:03):
So I personally think level five is unobtainable at this moment, but there are plenty of building blocks that many telcos have already put in place. Orchestration and of course the conversion to virtualised networks and cloud — that whole separation of the data layer and the network layer, which happened I guess about 10 years ago — has really started a revolution that has caused a ripple effect through the networks to allow that high degree of automation. And I think there's also the business process side. In this particular panel, we're not focusing particularly on the BSS piece, but the BSS piece is actually super important for the automation process because the order flows in. I'm actually working on that particular process right now. I'm acutely aware of that real disconnect between the order coming in and then the automation of setting up the network on the other side — and how, "Oh wait, we still do that by hand. Why?" So obviously looking at the automation tools is not enough.

(20:40):
It does really need to look at the workflows and make sure that — and there are obviously a few that probably should never be automated — the vast majority can be and should be automated. And that silo effect of "the business people do this and the network people do that" needs to be broken down. You really need to say there's no reason why we can't automate this. And I have that conversation with people every day.

Ray Le Maistre, TelecomTV (21:18):
Okay. George, the TM Forum must be looking at all of these building blocks and seeing what is mature enough to help automation occur in certain domains. Are there areas that are far more advanced — maybe orchestration platforms, or intent-based networking, which has had a lot of focus in the past few years and has a lot of platforms enabling it? What are the key building blocks that you would say are ready right now for network operators?

George Glass, TM Forum (21:53):
So we've certainly seen things like intent-based operations — which uses generative AI — as something that's coming to the fore, and people are starting to implement solutions where you describe the outcome as opposed to describing the technology you want to use to deliver it. Closed-loop controls, which use predictive AI — doing anomaly detection, automatic error correction, and feedback loops — those are mature areas that lots of our members are developing solutions around. Intelligent orchestration, where we're starting to see the use of agentic AI — that's still at the test phase, but that doesn't stop you starting to build solutions that use intelligent orchestration based on rules and data to deliver much more favourable outcomes. And the final thing I want to say is that really what we're seeing in terms of the overall transformation of the business — and what we're trying to drive through the TM Forum — is the concept of an autonomous enterprise.

(22:59):
The IT and the network are now both software engines within the organisation, and the skills needed to manage those two domains — which were separate for many years but are now coming closer and closer together — are actually converging. We need people who have a deep understanding of networks but also a real appreciation of software engineering, pattern matching, and abstraction. Merge those together, add in the data scientists and the AI capabilities, and you can start to build an autonomous enterprise. But you've got to think about it differently. This is not redesigning the processes you have today. This is taking a step back and reimagining: how do I deliver service to my customers using new capabilities, new technology, and the latest AI that can be embedded into the processes and the components — to deliver a ground-changing, earth-shattering experience?

(24:12):
And I'll talk about some of the numbers that support that later on.

Ray Le Maistre, TelecomTV (24:16):
Okay. So we're seeing some themes already here — the importance of bringing the different teams within operators together. Is that something that VodafoneThree has in mind? I mean, obviously you're a new company, so integrating two teams, but trying to do that really quickly. Have you come into this integration with all of this in mind?

Francesca Serravalle, VodafoneThree (24:37):
I think it's a very important point, because IT owns the enablers — the data lake, the data platform. So obviously the requirements come from business and network, and IT builds the enablement platform as well. So we're seeing more and more synergy. If I look at the key enablers in our journey, and if we agree that the definition of L3 is achieving domain autonomy and level four is cross-domain autonomy — I always like to simplify things, so I'm going to simplify five years of programme into a couple of points. The key enablers for level three are having a domain orchestrator to automate resource provisioning and network function provisioning, and then the AI hub that assures those resources and network functions, which can be localised within the domain. And I think the main challenge we have is: what do we automate first? Is it the RAN? Is it the transport? Is it the core?

(25:41):
That is the first question we need to ask for L3. Level four is end-to-end automation, cross-domain. So we need the service orchestrator to translate services into network policies across domains, and we need the federated AI agents and the orchestration that you were talking about. I'll leave intent aside because that touches a very complex area — it's more like next-generation BSS, where users access a category-driven marketplace, set intent, which is then translated to a service policy by the orchestrator. So that will be the key enabler. It sounds simple, but it's actually a very, very complex journey to navigate.

Ray Le Maistre, TelecomTV (26:30):
Okay. And we'll come back to domains and which ones to choose in a minute, but I also want to ask — in terms of the way that VodafoneThree is approaching this under the Vodafone Group — how much of this is being done in-house? Vodafone has been using the word TechCo for a while now as the way it's evolving as a company. How much of this development is happening in-house and how much are you working with suppliers that have always been on the roster?

Francesca Serravalle, VodafoneThree (27:07):
I think it's mainly a co-creation activity with partners, where we try to use our data science capability to drive MVPs using agile methodology and a squad that is multifunctional — data engineers, data scientists, and UX developers. So we have our own team to build and create this culture of innovating fast and failing fast, basically, like cloud providers do. But as I mentioned before, we also believe this is a multi-year transformation programme and we are working with partners to help us understand where we are in this journey. The role of partners is to help identify which use cases make most sense to prioritise, and what is the right value realisation framework we should use to decide what to scale. It's all about governance and framework — more like helping us adopt the right transformation journey. But then we leverage our in-house skills to drive execution speed and innovation.

Ray Le Maistre, TelecomTV (28:25):
Okay. Colin, are you seeing a lot more of this co-development and much tighter integration in the processes between HPE Networking and your customers?

Colin Evans, HPE Networking (28:36):
Absolutely. The co-development model is one we see very often. Going back to the earlier point about closed-loop automation being the thing that actually delivers what the intent suggests — the intent is created to say what the network is meant to do at a very high, abstracted level, and then closed-loop automation is the thing that actually delivers that and moves us from wherever we are — 2-point-something to 4-point-something. So closed-loop automation is extremely important. But the other point I'd make is the cross-domain aspect of this, because I think we sometimes think of domains as bits of the network — like the access layer, the core layer, the edge. But it also extends vertically. It starts with a service that starts on a container, which sits on a compute node, which is connected to a local network inside a data centre, which is then connected across the wide area.

(29:34):
So that cross-domain view should not just be in the horizontal sense of parts of the network, but also in the vertical sense of how the whole thing needs to orchestrate. And workloads are not static — they're defined from one point but they move around all the time, and we need automation capabilities across domains with closed-loop automation that follows those services as they move within the network. Equally, they could be delivered by multiple parties — so it can be on a multi-operator network as well in certain cases. That's equally important.

Ray Le Maistre, TelecomTV (30:12):
Okay. So key to all of this — and I think it's been mentioned — this isn't just a technology process for the sake of it. There are targets, there are goals, there are reasons for doing this. Anita, what kind of conversations happen within the NGMN about what can be achieved through automating different parts of the network and different processes?

Anita Döhler, NGMN Alliance (30:47):
I think it really depends on each individual operator with regards to how they prioritise. It can be energy efficiency, network efficiency in general, cost optimisation, MTTR. We have 17 operators, so we typically keep it more generic — to some extent that's natural, but to some extent we also need specific examples and best practices from operators. But I'd just like to echo what was said — one of the main outcomes of the recent work done on agentic AI-based operating models was that it needs cross-domain and AI-literate teams to collaborate and have a very clear way of measuring the outcome. Different use cases are listed there. But when it comes to priorities, I think it really depends heavily on the situation of the specific operator and the region they operate in.

Ray Le Maistre, TelecomTV (31:58):
So Francesca, do you have a sense of where you might start? You mentioned different domains — what are the key dynamics? Is it a business focus as much as anything, a return on investment based on particular parameters?

Francesca Serravalle, VodafoneThree (32:16):
There are a few key activities that we're going to drive, as we're going to invest £11 billion in the network and obviously want to do it in the best way. So we're going to introduce automation in network deployment — fibre and cell site deployment — automating the lifecycle of plan, design, and build through automation of HLD and LLD. And then during the build phase, you can do progress tracking, scheduling of work, and then automating the audit as well, using even video agentic AI. That is a high priority for us. And the second one is ticket resolution — it's a really complex activity because you need to leverage network data, service data, and customer data. You need to access network systems and customer systems. It's about linking network operations and service operations — OSS and BSS — and it takes a lot of time. So yes, the KPIs are mean time to resolve and number of calls to resolve, but I think the business outcome is better customer experience and operational efficiency — both of it.

Ray Le Maistre, TelecomTV (33:34):
Okay. And in a very measurable way — something that can be taken to, I guess, even regulatory filings to say this has been achieved. Would it go that far?

Francesca Serravalle, VodafoneThree (33:49):
I think it would be great. I can't say now how we will be able to track it, but it has to be. I think value realisation requires measurable business outcomes, and this is key for us to be able to scale. Scaling is the most difficult part — you can do an MVP in two months at small scale, but for scaling it's about embedding into the real business process, driving adoption, creating a proper blueprint architecture. It's not a beta product — it's a proper-scale deployment. So the value realisation and business outcome measurement is really, really key. Whether we will be able to bring it to the regulator or not, it has to be clear: what customer impact are we going to deliver, and what business outcome — in terms of cost reduction, service agility, or time to market — has to be a very clear set of metrics and KPIs.

Ray Le Maistre, TelecomTV (34:53):
And Beth, from your experience at Verizon — is there a specific network operation that gives a very clear and deliverable return on investment, one that needs to be top of the list and may already be delivering measurable returns?

Beth Cohen, Luth Computer (35:11):
For sure. I talked about the business workflows from quote to delivery — or quote to bill, as it's normally referred to in the industry. And at the very beginning, the biggest ROI is speed to provisioning, which is often the most expensive part because you have to do a truck roll, and you want to minimise that as much as possible. And so you can do a lot of things with AI — minimising the number of truck rolls you do, automating a whole bunch of stuff behind the scenes. So that's an area that really should get a lot of focus for return on investment. And at the other end is billing and SLAs, because SLAs can cost a lot of money and they tend to be expensive to manage because a lot of people are involved. Meeting your SLAs can obviously be great ROI, and not having to pay out credits is even higher ROI.

(36:41):
So I'd say the two ends of the process are probably where you get the most return on investment.

Ray Le Maistre, TelecomTV (36:48):
George, I'm sure you must have some stats to bring in here.

George Glass, TM Forum (36:54):
Yeah. So as I said, we divided the network into 55 scenarios. We then asked our members for those scenarios: A — did we have the technical solution that was capable and possible to implement today? And B — which of the scenarios would deliver the highest return on investment? Where was the most problematic area of your business? And from those 55 scenarios, we identified 22, which we called high-value scenarios. For those, we've developed technical solution packages, working with the members, for the patterns for how you implement those solutions. The highest-value scenarios, as voted by the members of the TM Forum, are — and for obvious reasons — fault management, because there's lots of manual activity and lots of customer dissatisfaction when you've got faults; service assurance; and energy efficiency. For fault management, it actually goes across all of the domains — IP, core, access, transport. And whenever you look at fault management in those domains — and I think it's been touched on already — it's not just constrained to the network, because obviously a fault on the network affects the customer experience.

(38:09):
It affects the engineer's workforce and their activities. And therefore, joining all of that up is actually where you get the transformation. And what we're actually seeing is members who have implemented solutions at scale. I had the pleasure of visiting the Guangdong region for China Mobile, where they have a solution that's been implemented for IP fault management. They've achieved a 50% reduction in mean time to repair. They've reduced their total number of faults raised by 15%. And they've automated the equivalent of 5,500 people's worth of activities with agentic AI. They've rolled this solution out to 300 million customers. This is not a proof of concept.

(39:14):
This is not a pilot. This is in production, in their NOC, delivering tangible business value to China Mobile. And I'm seeing that pattern repeated across Asia and starting to come into Europe as autonomous networks are being implemented following those patterns.

Ray Le Maistre, TelecomTV (39:36):
Okay. That's really, really interesting. It's great to have that use case and those stats. Is this something that can really make a difference? Obviously China Mobile has massive scale, but that means things can actually happen and make a difference. Colin, do those use cases tally with what you are hearing from your customers? What kind of metrics are they looking at? What are they hoping to get in terms of ROI?

Colin Evans, HPE Networking (40:08):
Well, I think the one that George just mentioned around fault management and fault remediation is probably the best one today that's potentially using the closed-loop model — where it's a real hands-off, virtually no-human-intervention type of model. So yeah, I would completely concur with that one. The others are around things like change management and change configuration in the network. A good example there is rolling out software updates — there are potentially tens of thousands of devices running code in these — let's not even call them networks, let's call them the cloud — because it also includes virtualised and containerised devices, servers, storage devices, and so on. So rolling out those software updates is another one that we very much see in terms of change management. And then the final one is obviously service provisioning. Once a service is defined, it's much easier to automate the provisioning of that service and again use a more hands-off approach.

Ray Le Maistre, TelecomTV (41:12):
Okay, excellent. So let's talk about fault management. This is my fault in terms of time management — I didn't factor in time for Q&A from the audience for this panel, because we've only got a couple of minutes left and we've got another session going live from the far side of the room in a couple of minutes. So I'm just going to ask our panellists for a quick takeaway thought to end the panel — what they would like to see in terms of what could be automated in 2026 and 2027. What should the priorities for network automation be in the next couple of years, starting with you, Beth?

Beth Cohen, Luth Computer (41:58):
Let's see. I'd say coordination across the millions of devices, actually, to keep them in sync and using AI to help drive that.

Ray Le Maistre, TelecomTV (42:13):
Okay, great. Colin?

Colin Evans, HPE Networking (42:15):
Just don't be constrained by supposed milestones that other people suggest you should be at. Do the things that make sense for your organisation, the things that deliver the most ROI, and take a step-by-step approach. If it takes several years, it takes several years — but try not to do it too quickly and make mistakes.

Ray Le Maistre, TelecomTV (42:35):
Okay. Anita?

Anita Döhler, NGMN Alliance (42:36):
Yeah, I tend to agree. I mean, everything we do is to generate customer value, and we need to have the customer value focus. And then depending on the biggest areas of impact for each individual operator, that's what needs to be tackled first.

Ray Le Maistre, TelecomTV (42:52):
George?

George Glass, TM Forum (42:55):
You don't have to build from level two to level three to level four. You can actually jump from level two to level four if you follow the technical solution packages we are publishing with our members.

Ray Le Maistre, TelecomTV (43:09):
And Francesca?

Francesca Serravalle, VodafoneThree (43:11):
We have built the enablers. We have network, customer service, and data centralised. It's now about starting to correlate those data to have demand-led network expansion, user-led network optimisation, and a fault management that correlates — as you were saying — customer data and network data, OSS and BSS data, basically.

Ray Le Maistre, TelecomTV (43:33):
Okay, excellent. Well, thank you very much to our panellists. We do have to wrap up there — that's all we've got time for for this session. We have a break now of about 30 minutes. We're back at 11:15 AM UK time for the final morning sessions. For those watching online, the Extra Shot programme with Tony Poulos is coming up in just a matter of minutes. For those here in Windsor, the pinball tournament resumes — don't forget that. The coffee cart is open again and nibbles are being served out in the foyer. Any questions you had for our panel, please grab them during that coffee break. Let's have a round of applause for our panellists. Thank you.

Please note that video transcripts are provided for reference only – content may vary from the published video or contain inaccuracies.

Panel Discussion

A comprehensive panel discussion reveals the stark reality of network automation progress, with most operators scoring just 1.5-1.8 on the TM Forum's autonomy scale despite ambitious level four targets. Industry experts from Vodafone, HPE, NGMN Alliance and TM Forum examine the gap between zero-touch aspirations and operational reality, discussing practical approaches to bridge this divide through business-led transformation programmes, data platform investments, and the evolution from machine learning to agentic AI capabilities.

Broadcast live 19 May 2026

Participants

CO-HOST

Francesca Serravalle

Head of Monetisation, VodafoneThree

Anita Döhler

Chief Executive Officer, NGMN Alliance

Beth Cohen

Telco Industry Analyst Luth Computer

Colin Evans

Snr. Director Business Development, HPE Networking

George Glass

Chief Technology Officer, TM Forum