New predictive networks allow CSPs to stay one step ahead

Commercial context

Before launching new services, CSPs must first manage the messy complexity underneath. Legacy OSS systems are siloed and expensive. Data is fragmented across RAN, core, edge, and IT layers. Planning and maintenance become reactive, energy use inefficient, and troubleshooting still relies on manual workarounds and truck rolls. Inventory data is often spread across incompatible systems. All of this makes failure harder to predict — and ultimately, this fragmentation blocks AI adoption and real-time decision-making. The cost? Missed SLAs, rising OPEX, and slower response to outages — all while customer expectations keep rising.

The ‘Predictive Intelligence for Optimized Networks & Enhanced Experience Resilience (PIONEER)’ Catalyst, led by Globe in partnership with Singtel Group and Dell Technologies, is designed to remedy this issue. In doing so, it provides a working, AI-driven architecture to unify data, reduce downtime, and guide CSPs toward full autonomy — all grounded in TM Forum standards and known operational needs.

The solution

The project team recognized that fractured, inconsistent operational systems prevent CSPs from building truly autonomous, AI-ready networks. So, rather than layering automation on top of legacy silos, the team designed a new architecture — one that combines AI intelligence, real-time orchestration, flexible cloud infrastructure, and unified data — all aligned to TM Forum standards. The engine works closely with digital twin models that simulate network behavior in real time. Before applying any optimization, the system tests it virtually, reducing the risk of misconfiguration or unintended side effects.

To act on insights without delay, the architecture includes a broker layer — a real-time coordination system that connects domains through event buses and decision engines. It processes over 3 million records per second and triggers orchestration decisions with minimal latency. When the system detects a critical issue in the RAN or edge domain, it immediately initiates the appropriate response — whether that’s rerouting traffic, reallocating compute resources, or notifying field teams in advance.

Supporting this intelligence is a common telco cloud platform. Built using Kubernetes and open-source standards, it provides the computing foundation for both traditional network functions and AI workloads. The platform is flexible enough to support GPU acceleration, and consistent enough to deploy across core, transport, and edge domains without vendor lock-in. It allows CSPs to scale AI functions without rebuilding their infrastructure from scratch.

Crucially, none of this works without trustworthy data. That’s where the data normalization layer comes in. This component aggregates structured and unstructured data from across the network — telemetry, alarms, logs, experience data, even business inputs — and transforms it into a unified, machine-readable format. Aligned with TM Forum SID standards, the data layer becomes a single source of truth for operations, planning, and customer experience. It also supports both real-time streaming and batch analytics, unlocking predictive insights while cutting long-term storage costs. Taken together, these components form a foundation for what comes next in network operations. Not just automation — but anticipation. Not just fixes — but foresight.

Wider application and value

As Dennis R Abella, VP and Head of Network Digitalization at Globe Telecom, put it: "Globe isn’t just testing theory — it’s building a blueprint others can follow. As the lead OpCo in the Singtel Group’s Autonomous Networks program, Globe is validating L3 and L3+ use cases today, with L4 deployment targeted for 2025. The aim isn’t a one-off demo — it’s a production-ready model that can scale across the group."

And the results speak for themselves. Predictive assurance and closed-loop automation have reduced mean-time-to-resolve by up to 90%. Smarter infrastructure decisions and automated planning have already cut total cost of ownership by more than 30%. Service reliability has climbed alongside 95% data accuracy, while customer experience has seen an 80% improvement. What makes PIONEER stand out is its emphasis on transition — not experimentation. The team is working with live network data, using open APIs, and building against real-world operating conditions.

The impact of the solution could be significant. A standardized, AI-ready OSS model paves the way for GenAI in production — from LLM-assisted operations to predictive maintenance and intent-based service design. Unified, machine-readable data unlocks modeling and forecasting tools that were previously out of reach. And with AI-optimized networks comes lower energy consumption, smarter resource allocation, and a more sustainable digital backbone. Taken together, this has major implications for service development, enabling CSPs to move from asking “can it work?” to “how soon can we launch it?”

But perhaps the most meaningful outcome is resilience. Not just against faults — but against the pace of change itself. When demand surges, weather hits, or critical infrastructure fails, networks need to adapt fast. PIONEER is helping build systems that don’t just react — they think ahead. PIONEER is a step toward that kind of network. One that doesn’t just continue operating, but stays ready.