CxO Insight

Network automation’s role in balancing cost to serve with future benefits

Axiata Group’s Thomas Hundt explains to Contributing Editor Annie Turner explains that network automation’s role in balancing cost to serve with future benefits is a delicate balance, buffeted by macroeconomic forces, and timing is key.

Thomas Hundt is Group Chief Strategy & Technology Officer at Axiata Group, based in Kuala Lumpur, Malaysia. He has held senior roles within the group since 2008, which has mobile operations in Malaysia (celcomdigi), Indonesia (XL Axiata), Bangladesh (Robi Axiata), Nepal (Ncell Axiata), Sri Lanka (Dialog Axiata) and Cambodia (Smart Axiata) of which Hundt has been the CEO from its inception in 2008 till 2021.

It also runs an infrastructure business, edotco, which has a portfolio of more than 58,000 towers across Malaysia, Myanmar, Bangladesh, Cambodia, Sri Lanka, Pakistan, Indonesia, the Philippines and Laos. It has an AdTech business, ADA, and a FinTech business, Boost.

The newly merged and jointly held operator company in Malaysia, celcomdigi, was created by Axiata combining its Malaysian opco with that of Telenor in November 2022. Also, towards the end of last year, the new Malaysian government decided to review the previous government’s 5G policy of building an open, single wholesale 5G network, Digital Nasional Berhad (DNB), to be used by all the country’s mobile service providers. As the situation is in flux, our interview focuses on the Indonesian, Bangladeshi, Nepalese, Sri Lankan and Cambodian markets.

These countries are culturally diverse, but to a greater or lesser extent are all low-affordability markets, with ARPU of between $5 and $2 at the bottom end of the range. All have and are still suffering from the effects of the pandemic, then the cost-of-living crisis, with Sri Lanka hit particularly hard. At the same time, as Hundt observes, “The operating environment is very intense capital-wise and that drives our thinking about operational efficiency: cost-to-serve is probably the primary concern.”


Cost to operate with an eye on the future

So far, Axiata has only launched 5G in Malaysia through DNB but has shut down its 3G networks in most markets. Hundt stresses, “4G is our bread and butter [but] we are not behind the curve in terms of having state-of-the-art networks because only by going for state-of-the-art networks can you hit the cost positions we are striving for.” He adds, “We are looking at our cost structures and network design from an end state where 100% of revenues are coming from data-driven services.

He also points out that 5G is about more than the network: it relies on the business case, regulatory readiness and ecosystem development, availability of 5G-ready devices and their affordability, and the availability of spectrum. “Any of these components can delay progress in markets,” he says, adding that he expects some solid steps towards 5G deployments next year in some of Axiata’s markets.

Hundt notes the timing of the 5G investment cycle in the wider industry is “not necessarily translating immediately into returns”. Hence he expects 5G’s more efficient use of spectrum will be the biggest driver initially because as 4G networks hit their capacity ceiling, “It will not be economically viable to fix a capacity crunch by adding more and more sites – 5G is needed to widen the pipe for eMBB [enhanced mobile broadband],” he explains.

So, although Axiata is very interested in the potential of private networks, massive IoT and other innovative B2B use cases, “none will fly or be a financially solid investment case unless the consumer ecosystem adopts 5G,” he argues.

“When we look at designing our network and IT stack, costs to operate, synergies and efficiencies are against that background.” He adds, “Energy costs and carbon footprint are… of high priority. This includes the cost of building reliable networks in markets where the stability of the grid varies.”


Consuming energy

Hundt explains, “We are tackling [energy consumption] on multiple fronts. The first is by optimising infrastructure by investing in modern power systems. We have mostly abandoned cooling at radio sites by transforming legacy indoor sites to outdoor. We have put solar panels at thousands of sites across all our markets, wherever they make economic sense so, to have at least a hybrid power supply. In parallel we focus on the active radio components where we are working on AI and power algorithms to kick in,.”

“We have tested several energy optimisation solutions. Everyone touts them but you only find out which make dynamic impacts by testing. Results depend on traffic profile, and the patterns you’re running. The first wave of energy efficiency solutions relied on simple manual triggers.” Hundt is hoping for greater sophistication but says so far study outcomes “are not up to our liking” and that “the real impact is yet to come”.

“I believe if full machine learning/AI algorithms were deployed to make dynamic changes, then we could see significantly better results. Of course, network modernisation plays its role as well. If you run outdated gear, you shouldn’t be surprised that energy consumption is very high – although we need to manage modernisation capex as well. Energy consumption is a very broad topic and a constant development effort with our partners.”

After cost to serve, Hundt says networks’ reliability and availability are Axiata Group’s main concerns and “the complexity of networks and operationalising them is why network automation and autonomous networks are very, very important”.

“Achieving [ETSI’s] Autonomous Network (AN) Level 3 by 2025 is our goal. However, it’s not just box ticking to say, ‘We’ve automated at this level,’ but about what is really driving our operational efficiency, and which use cases can make a difference, built on an AN approach. That’s where analytics and automation at scale come into play. We have dedicated teams including what we call an AI factory to help operationalise it.”

Hundt continues, “This is a framework, a concept that includes common architecture and tooling, use case and algorithm sharing, talent development etc. We have deployed ‘AI factories’ across our opcos. Further, we operationalise network and IT across our opcos through an organisational structure called the ‘Collective Brain.’ We have a relatively light team at group headquarters – a handful of people – then very smart brains out there in the opcos. They solve common problems together.”


Being data driven

“The tools we are using, how we operationalise and deploy them into the network and IT stacks, everything starts from collecting, cleansing, retaining, growing and governing data; the architecture of the data lake and so on is our own,” Hundt adds. He acknowledges, “The biggest challenge still, frankly, is data governance, data collection, data retention – assembling all sorts of network and IT data – because garbage in, garbage out. If you haven’t built a very robust engine underneath you will not get to the promised outcomes further down the line.”

For now, he says, “I believe network automation really kicks in when it comes to the best way of sweating assets. How can we maximise Quality of Service and capacity? We need self-optimised networks (SON) tuning the network constantly, then things like fault prevention and preventive maintenance so that up times are optimised. These are the places we’ve seen automation work best so far with good but not satisfactory results yet.”

Hundt states: “I believe that, especially in context of network optimisation, our industry is behind the innovation curve. Solutions out there are still primarily static, trigger based and set to predefined parameters. There’s not enough machine learning capability in the network domain to constantly iterate and improve itself, so that’s a major opportunity for development.”

He is upbeat about network automation’s role in network and capacity planning, Axiata is investing in solutions to allocate capex for network expansion capacity more optimally. “Those are the areas in which we are putting probably our biggest effort and we anticipate seeing good results,” he says. “When it comes to network operations, virtual NOCs, we are yet to see the outcomes but we are working on developing and deploying use cases along those lines, as per our aspiration of hitting Autonomous Network level 3 by 2025.”


What does transformation mean?

He notes wryly that the term ‘transformation’ always seems to be used “with a flavour of a timely, limited process being kicked off”. This doesn’t fit his view: “Transforming means you’re never standing still,” Hundt says. There is no goalpost – they move permanently, so it’s a never ending thing. From a strategy standpoint, you need sufficient angles of attack, and that obviously includes strategic considerations such as delayering and a lighter than the traditional full-stack approach we’ve had over decades.”

He states, “To achieve all this delayering we must adopt progressive ways of thinking and working…At the same time, certain fundamentals remain. Connectivity has its place, and perhaps even more so than [was obvious] a few years back, but it has to be done via the most efficient and effective setup, which we are still trying to figure out. This is constantly the focus of operational efficiency and lowest cost to serve. We have a place there. The question to telcos is, ‘In which layers do you want to play and have the right to win?’.”

Hundt does not see “moving up the value stack” as necessarily “the most optimal move”, but clearly Axiata Group is keeping its options open. It was only the second telco in the world to be certified as ‘Running on ODA’, that is TM Forum’s Open Digital Architecture compliance certification. Hundt says, “We’re very much engaged with GSMA on Open Gateway and CAMARA which are great initiatives.”