Roderick (Eric) Santiago is FVP and Deputy Network Head at PLDT. Here he explains his automation and network strategy, which starts “CIA”, to Contributing Editor Annie Turner

PLDT is the largest integrated telco in the Philippines, offering fixed and mobile services. It has the most extensive fibre footprint in the Philippines with homes that reach 18.13 million people in 71% of the country’s towns and 91% of the provinces. The combined 4G/5G network of its wireless arm, Smart, covers around 97% of the material population.

Eric Santiago has been FVP and Deputy Network Head at PLDT since January 2022 and, having lived in the US for 22 years, likes to describe his strategy as “CIA”. He says, “The ‘C’ stands for cloudification. There is an opportunity to cloudify our orchestration platforms to virtualize network functions, enabling more scalable and automated operations across the network.”

The “I” stands for innovation: delivering end-to-end, innovative products by monitoring orchestration. “We have numerous innovative tools that will be incorporated into the OSS modernisation programme and the BSS,” he continues. The OSS modernisation is central to Santiago’s strategy; it began in the first half of 2025 and is slated for completion by the end of this year, with AIOps at its core. “We’re integrating APIs into the network and applying robotic process automation,” he notes, “as new technologies continue to be deployed.

“The ‘A’ in CIA is automation, but its success depends on building the company’s people capabilities. My goal is to develop those capabilities as part of the automation effort.”

Foundations in place for CIA

There are already solid foundations for the CIA goals. PLDT has maintained a self-optimising network for the past three years, including automated power-saving. Parts of the mobile network are placed in sleep mode when traffic is low – such as cell sites or in-building solutions in malls – reducing power consumption and costs.

The operator also runs a self-optimising network for disaster recovery and other optimisation segments. “If there’s an outage, we can automatically detect and fix it dynamically,” Santiago says.

PLDT is conducting many proofs of concept (PoCs). One focuses on auto-detection to take a proactive approach to network congestion. “The aim is to reallocate traffic dynamically, moving it to unaffected areas during calamities; Philippines experiences around 20 typhoons a year,” he states. “We need to leverage automation to improve resiliency, restore outages faster and deliver better quality of service.”

PLDT is also upgrading much of its passive fibre broadband infrastructure, deployed in 2011, by adding active equipment or reflectors. This enables fault detection and precise identification of where the fibre is cut or where the fault occurred, so issues can be fixed quickly. A learning mechanism will dynamically notify engineers or technicians assigned to the area so that fixes can be completed before customers notice. “One PoC is experimenting with AI to detect, diagnose, and resolve issues before dispatching a technician,” Santiago adds.

Much of the automation described relies on closed loops, but PLDT also employs zero-touch provisioning for SIM cards, service delivery and activation that runs without manual intervention and is part of a broader order-management automation.

Other PoCs explore agentic AI and GenAI within the OSS. PLDT maintains partnerships with NTT DATA and collaborates with Microsoft and Google, and there has been progress recently working with AWS to embed AI into operations.

Scaling up without snags

It is often hard to translate PoCs into at-scale deployments. How is Santiago approaching this issue? “We will deploy a phased approach: incremental integration with current systems and milestone-driven progress on the OSS modernisation platform. Upon finishing a release, for example, in customer service, we will overlay agentic AI.

“This approach is a multi-year programme designed to ensure that billing, revenue assurance, and fraud detection capabilities are simultaneously monitored and protected against AI-enabled fraud.” Already fraudsters have begun leveraging diverse AI techniques, underscoring the need for robust governance.

Autonomous network journey

How does Santiago gauge PLDT’s overall progress in automation? Citing TM Forum’s Autonomous Networks framework, he places PLDT between Level 2 and 2.5. “For me, Levels 4 and 5 are essentially a virtual NOC,” he says, “which is why I position ours at 2 to 2.5 currently. The OSS modernisation, however, should unlock further improvements as we add more predictive analytics to address customer faults, churn, and demand forecasting.”

He explains that automation levels vary by function and domain, and depend on the intended benefits. “In telco, our top priority is reducing operational costs, followed by faster service delivery and an enhanced customer experience,” he says. Santiago also emphasises reliability and resiliency. “These are the top four priorities I’m focused on. To rate progress above Level 3 and toward Level 5, I want to ensure we meet all the KPIs.”

What will PLDT look like in 2030?

By 2030, Santiago expects PLDT will have fully automated provisioning for network management and operations. He envisions a self-healing network that resolves issues with minimal human intervention and dynamic optimisation that allows traffic to be rerouted to improve performance more quickly. A robust fault-monitoring system would alert engineers and enable fixes before outages occur.

On the customer-service front, he foresees more virtual assistants to resolve common issues and a unified, automated ticketing system that can fix problems and escalate them as needed. He is already advancing delivery, assurance and activation, with a goal of zero-touch provisioning and total automated management by 2030, integrated with existing systems.

Santiago acknowledges integration is a major challenge: “We’re working on APIs or middleware to connect automation tools with our OSS and BSS, but in the new OSS you’ll have CRM and NMS tools. How you integrate them more efficiently is critical.”

He also envisions near-total national coverage by the end of the decade, reaching the remaining 3% that currently are without 4G and providing fallback mechanisms for those with coverage gaps. This will be achieved through a mixture of infrastructure, macro towers, small cells, fixed wireless access, broadband wireless access, satellite and microwave, tailored to local conditions. “The main constraint is getting transport and power in place,” he says.

PLDT is already deploying alternative transport technologies arising from completed PoCs. “For example, instead of relying solely on fibre, repeaters or microwave links, we’ve tested laser technology in partnership with Google. We’ve deployed it in the network and it’s reduced costs and increased reliability versus microwave,” Santiago elaborates, adding “By 2030, satellite connectivity (direct-to-cell technology) should also be in place to further boost resiliency.”

This diversified transport mesh will enable scenarios such as home routers automatically falling back from fibre to mobile or satellite in case of failures.

AI in the 2030 network

“By then, we will deploying, embedding AI everywhere in the network…but we need the discipline to use it properly. We need to continuously work hard and think strategically about how to improve things further,” he muses. “I think by 2030 there will be more robots assisting us as well and better trained large language models (LLMs) will eliminate some of the biases that I have seen in some of the languages we have.”

Small language models will be important for specific tasks and use cases. “That’s something we have learned as well – you don’t need always to develop big ones. You could start small, deploy quickly, and monetize where possible,” he states. “The key moving forward will be our ability to identify the right use cases, then choose the tools, design and map the processes, and develop it accordingly to beat the competition.”

Santiago concludes, “It’s important we train models accurately with clean data, which is another reason for the OSS modernisation. PLDT is a 98 year old company. By the time we turn 100 in 2028 we want to have highly clean and reliable data. That clean data will underpin all large language models (LLMs) and the small-language models we’ll build. As I say, AI is garbage in, garbage out, so we’re deliberate about data quality from the start.”