Ken Tan, CTO of Malysia’s DNB, talks about the importance of a strong data foundation in readiness for AI, leveraging data across the organisation to drive automation for operational efficiency and enhance productivity.

Part of the CTO Interviews series featuring Ericsson customers sharing insight into the journey to Autonomous Network Operations. What is the key to building autonomous networks?  Get the details from Ericsson.

Digital Nasional Berhad (DNB) was established in March 2021 as a 5G wholesale infrastructure to six mobile retail service providers. Ericsson was chosen as the network equipment vendor through a tender process. The 5G network was built within nine months and began offering 5G service in December 2021.

One of DNB’s core principles is to lower the cost of connectivity to bridge the digital divide. The operator applies a digital operating model based on AI and algorithms to guide the most critical processes and operating decisions and deliver value to customers. DNB’s CTO, Ken Tan, says DNB’s initial approach to automation was embedding data into processes, which he describes as Level 2 in TM Forum’s model for autonomous networks.

Now DNB has reached Level 3, for example, reducing alarms from 300,000 per day across 500 sites to fewer than 1,000 alarms per day for 7,500 sites. “I can scale up effectively without increasing the cost to the company, allowing connectivity to become a commodity,” he says.

Data ready to leverage AI

Tan adds, “If you’ve got the right and trusted data available, Gen AI will unlock the potential for more efficiency in the network”. For example, AI can leverage more granular, real-time data about the 5G network’s use to automate power management in real time, adapting it to traffic levels. This new method of operating DNB’s 5G network yields up to 8% better energy efficiency which translates into millions of ringgits in savings.

Enterprise processes can benefit as well as operations, Tan states. For example, all DNB’s paper-based site contracts have been digitalised using large language models (LLM) to ensure current terms and costs are applied to all contracts. It also triggers timely, automated payments to landlords. This has improved efficiency between departments, removes opportunities for fraud and provides real-time insights. The insights enable cost comparisons between available sites in the market and can be used to optimise cost.

Sharing and learning

As a strategic partner, DNB is leveraging Ericsson’s global leadership in 5G to drive AI innovation and economies of scale. Tan says that DNB has been applying this in the domain of cybersecurity where “AI is trained to check any anomalies in DNB’s cybersecurity postures and controls”. Instead of hiring cybersecurity consultants, DNB benefits from what Ericsson learns across its global customer base, which mean the vendor can develop AI use cases for cybersecurity to protect networks.

This sharing allows DNB and Ericsson’s other customers to secure against new cybersecurity vulnerabilities at very low cost, according to Tan.

Automation for the right reason

He continues, “This ties back to doing the right automation for the right reason. Do not automate problems. For example, if a customer’s order is failing to complete and automation retries the order, it will probably get through, but without fixing the root cause. Automating problems can compound the issue in the network, similar to making a ‘spaghetti mesh’ that is so complex it is very hard to untangle. Organisations typically require the costly engagement of consultants or external help to fix it.”

He goes on to explain that automation should be included in the upfront planning process. Tan stresses, “Do not automate at the end. Anytime you make a change, plan for automation and apply it, especially in the testing phase. DNB’s radio network is connected to six different core networks. When a DNB engineer makes a change for say, operator one, it must also be tested for each of the others to ensure it doesn’t introduce a fault into any of the core networks.”

If done correctly, automation can be applied to all six networks rather than each one having to be tested individually, requiring six teams with six lots of devices. This not only delivers savings, speed and efficiency, but Tan says, “becomes the foundation to build the autonomous network”.

He continues, “There’s always a certain efficiency level at which you need to start investment in capabilities. I’m still asking, ‘How do I maximise the existing capability?’. I have to automate as much as I can.

Self-healing is next on Tan’s automation agenda. “We can automate some simple stuff based on our data that we are happy to hand-off,” he adds. In these cases, a network fault might just need the rebooting of certain components, which can be done quickly, without human intervention. “This is how my engineers are evolving from DevOps to AIOps,” he explains. “I can’t find the justification yet for a common AI engine to introduce more complex self-healing use cases because today all the LLMs or engines are unique – you cannot port from one to another. It’s like if you’re using Google Cloud and you shift to Microsoft cloud, the cost of migration is prohibitive.”

So the big question is, Tan says, “Can we commercialise Level 4, Level 5 [of the TM Forum model]? We’re going through that process to understand the benefits versus the trade-off in effort.”

Close to intent-based operations

Ericsson is part of that equation because it has the scale, scope and expertise to invest in R&D where it can most help customers, now and in future. “I use Ericsson’s AI engine for intent-based operations,” he says. These operations have been in incubation for about three years and are close to being ready for commercial use.

“It will innovate the way we maintain the network,” Tan states. For instance, in future, he will be able to issue verbal instructions to the network to close down bases for maintenance. “The network will use data based on the OSS to ensure the surrounding sites can take care of the traffic load. The key to that process is what to do if something goes wrong,” he continues.

“The big danger in using this technology is if you don’t have the data to tell you whether it’s right or wrong. Because of unknown unknowns, you take steps. Data goes through many iterations making many unknown unknowns known. This is where you mitigate risk and innovate.”

Bold steps to win with 5G

Tan says innovation in 5G is about bold steps – using AI and automation to make 5G a commodity for consumers but involves altering the whole cost structure for applications like Industry 4.0, smart cities and IoT. He adds, “We need these automations to work out, as an industry. The question is how to drive towards that aim?.

Tan thinks industry-wide collaboration is the answer and points to TM Forum having common standards that enables the industry to collaborate, using the same terminology. In his view, the big issues to be addressed are latency, saving energy and enabling services like reduced capability (RedCap) for IoT. In this context, Tan says the central issue is, “What value are you adding to the enterprise through differentiated services that they can then deliver to the end user?”.

Experimentation and experience are the only way to tackle the “unknown unknowns,” Tan says, citing how DNB supported a national broadcaster to deliver multilingual live stream for the country’s National Day. Previously, broadcast trucks were always needed for outside broadcasts which are very costly to set up and limit the coverage area. With 5G, DNB could guarantee streaming video at 20% of the cost of the old set up which paves the way for the broadcaster to introduce AI in its future streaming service.

Tan explained, “This broadcast set-up was a challenge but now it’s done, it can be replicated elsewhere. Now we need to continuously find challenges for which we can find with commercial solution. This is where DNB can incubate with Ericsson and leverage Ericsson’s global customer base to scale. Successful incubations allow us to replicate a commercial model where we learn from each other what works, what doesn’t work what will drive growth in the long run.”

He concludes, “Otherwise, it’s going to be a lose, lose situation for us all; if there’s no revenue to drive investment, just high capex – it’s a downhill spiral. If we can grow that top line, it’s a different story.”