Photo by Karl Solano on Pexels
For years, the telecommunications industry has wrestled with a familiar question: how quickly can we adopt artificial intelligence? But according to Globe Telecom’s top executive, that question is now the wrong one to be asking. Speaking before a packed audience at MWC Shanghai 2026, Globe president and chief executive officer Carl Cruz reframed the conversation entirely — the industry has largely embraced AI in principle, he argued, but turning that embrace into measurable, scalable results is where carriers are genuinely struggling.
“Adoption was never really the hard part,” Cruz said during his keynote address. “Every operator in this room has a slide deck with an AI strategy. The hard part is making it work — at scale, in production, in real network environments where the stakes are real.”
From Strategy Slides to Operational Reality
Cruz’s remarks reflect a growing sentiment among senior telecom leaders globally. After years of heavy investment in AI pilots, proof-of-concept deployments, and vendor partnerships, the industry is entering what analysts are calling the “execution era” — a phase defined not by experimentation, but by the ability to operationalize AI across billing systems, network operations centers, customer experience platforms, and infrastructure planning tools.
Globe, which serves more than 90 million subscribers across the Philippines and has been one of Southeast Asia’s more aggressive adopters of digital transformation, has itself encountered the friction points Cruz described. The carrier has deployed AI-driven tools across predictive network maintenance, dynamic spectrum allocation, and customer churn modeling — but Cruz was candid about the organizational and technical complexity involved in scaling these capabilities beyond isolated use cases.
“You can run a successful AI pilot with a team of ten engineers and a clean dataset,” Cruz noted. “Deploying that same model across thousands of base stations, integrating it with legacy OSS/BSS systems, and keeping it accurate over time — that’s an entirely different engineering challenge.”
The Three Pillars of AI Execution Failure
Cruz outlined what he sees as the three primary reasons AI initiatives fail to move from pilot to production within telecom organizations:
1. Data Infrastructure Gaps
Telecoms generate enormous volumes of data — call detail records, network telemetry, customer interaction logs — but this data is frequently siloed across organizational units and stored in incompatible formats. Without a unified, real-time data fabric, even the most sophisticated AI models produce unreliable outputs. Cruz emphasized that data modernization, not model selection, is often the foundational blocker for AI at scale.
2. Workforce Readiness
The talent dimension is equally critical. Cruz pointed out that while telecoms have been hiring data scientists and ML engineers, the broader workforce — network technicians, customer service agents, field operations staff — must be upskilled to work alongside AI-augmented tools. “AI doesn’t replace your people overnight,” he said. “But it does require your people to change how they work, and that cultural shift takes sustained leadership commitment.”
3. Legacy System Integration
Perhaps the most technically complex barrier is the integration of AI capabilities with decades-old operational support systems. Many global carriers still run BSS/OSS infrastructure built in the 2000s, creating significant API compatibility, latency, and data pipeline challenges when attempting to inject modern AI inference engines into live network workflows.
Globe’s Own Execution Playbook
Despite the challenges, Globe has achieved notable wins. The company has deployed AI-based anomaly detection across its 5G SA (standalone) core, enabling network operations teams to identify and respond to service-affecting events significantly faster than traditional rule-based systems allowed. In customer operations, AI-powered virtual assistants now handle a substantial portion of tier-one support interactions, with continuous model retraining based on live conversation data improving resolution accuracy quarter over quarter.
Cruz also highlighted Globe’s investment in internal AI governance frameworks — a layer of oversight that many carriers overlook. Responsible AI deployment, he argued, is not just an ethical imperative but a business one, particularly as regulators across Asia-Pacific begin introducing requirements around algorithmic transparency and automated decision-making in consumer-facing applications.
A Signal to the Broader Industry
Cruz’s keynote arrives at a pivotal moment. Globally, telecom operators are under simultaneous pressure to reduce operating expenditures, grow digital revenue streams, and deliver on the network performance promises associated with 5G rollouts. AI is increasingly positioned as the mechanism to achieve all three — but only if operators can move past theoretical frameworks into functional, production-grade deployments.
Industry analysts have echoed this view. According to recent research from Analysys Mason, fewer than 30% of telecom AI initiatives progress beyond the pilot stage, with integration complexity and unclear ROI metrics cited as the leading failure factors. The gap between AI ambition and AI achievement, it appears, is industry-wide.
Looking Ahead: Execution as Competitive Differentiation
As 5G-Advanced networks mature and the early contours of 6G research begin shaping vendor roadmaps, the operators that master AI execution — not just AI adoption — are likely to emerge as the structural winners of the next decade. The ability to automate network optimization, personalize service delivery at scale, and predict infrastructure failures before they affect subscribers will define service quality in ways that raw spectrum holdings or tower counts simply cannot.
Cruz closed his keynote with a message that resonated well beyond the Globe brand: “The operators who win the next ten years won’t be the ones with the best AI strategy document. They’ll be the ones who figured out how to actually run AI in production — reliably, responsibly, and at the speed the network demands.”
In a room full of executives who have spent years building those strategy documents, it was a timely — and perhaps uncomfortable — reminder of how much work still lies ahead.
