- Dialog Axiata has now started using AI for unstructured data after leveraging AI for structured data for nearly a decade
- Dialog is also on its way to launch commercial 5G services, which would help in addressing the increased AI-based workload
- Key hurdles include trust, cost and process alignment
Dialog Axiata, Sri Lanka’s largest service provider, has been running AI-based applications for nearly a decade. But it's recently tried something new: tackling advanced AI use cases that leverage unstructured data.
“Over the last year and a half, we have started working with AI on unstructured data, using it for knowledge bots and chatbots," Dialog Axiata’s Group Chief Analytics and AI officer Romesh Ranawana told Fierce Network on the sidelines of FutureNet Asia 2025. "We are now exploring more advanced use cases, such as marketing and agentic automation of campaigns."
Ranawana explained there's a growing amount of unstructured data coming in from sources like audio logs, images and videos. He noted the operator is still grappling with how to properly manage this trove of information while also controlling costs.
"A few GenAI use cases that we have built are built in silos, but going forward, we can’t have different silos of data," he said. "There is a need for a more structured architecture where this unstructured data is coming in. And then there is a machine learning processing layer, which then extracts insights out of it."
Foundation to automation
Ranawana claims that Dialog Axiata has been using AI for nearly a decade now, with primary focus on structured data for tasks like churn prediction, upselling, cross-selling and forecasting, among others. “These AI models are already in production and integrated into core business operations,” he said
While it has long been using machine learning and AI, the operator is still some time away from end-to-end automation. The latter, Ranawana said, is a work in progress.
“Especially on the GenAI side, the big factor is trust, how to build trust among employees," Ranawana said. "We see this as a phased approach: first, using GenAI to provide knowledge so staff can quickly get answers; second, augmenting employees by automating parts of processes while keeping control with the business; and finally, once trust is established, moving towards deeper automation. Full end-to-end automation may still be a year or two away—it’s too early to go there now."
Dialog has around 17 million subscribers across mobile, broadband and fixed services. Axiata Group has also initiated the group-wide A3 (Axiata AN and AI) strategy to use AI and autonomous networking to transform its networks for greater programmability, flexibility, scalability and openness. Dialog Axiata is part of the Axiata Group, which operates in several markets in Southeast and South Asia.
Learnings from AI experience
Ranawana believes that selecting the right use case and evaluating the cost of AI rollouts are the key challenges in AI deployment.
“Our learning has been that it is important to focus on one or two high-impact areas instead of doing several use cases at the same time. Cost is a significant factor, as is picking the right use cases,” said Ranawana.
“The biggest challenge isn’t really technical, it is operational. If you start with technology and then try to change people and processes, it becomes an uphill task. A more effective approach is to first identify and redefine the process, determine the people who will operate within it, and then build the technology. So, it’s really reversing the traditional approach… It’s not a technology-led approach but a process-led transformation. That has been our key learning," he added.
Earlier this year, Dialog Axiata announced the completion of the “AI-for-Network” trial in collaboration with tech giant Meta to improve network optimization and Quality of Experience (QoE). It used Meta’s Llama AI models to analyze and optimize Dialog’s network to improve network performance. Following this trial, the service provider plans to roll out the deployment of the solution across the country.
Another challenge in AI adoption is telco network readiness, as AI traffic continues to increase.
Ranawana was candid that the networks are not yet ready to deal with the surge in AI-related traffic. However, he believes that the upcoming launch of 5G will help them address this to an extent.
Dialog Axiata is likely to launch commercial 5G services by the end of this year. Earlier this year, it completed a trial of Voice over New Radio (VoNR) on a live 5G network.