Close to the Edge: AI’s Expanding Role in Telecom Networks
AI is often hyped in telecom conversations, but its real value lies in clear, targeted application. In a recent FutureNet World interview, Benjamin Hickey, Director of Product Portfolio Management & M&A, Software Networking at IBM, offered a grounded view on where AI—and specifically automation, generative AI (GenAI), and advanced analytics—can transform networks if approached strategically.
Edge Computing: Decentralization with Purpose
The industry has long swung between centralization and decentralization. While public cloud brought an era of centralization, Hickey argues that decentralization — especially for AI inference — will gain traction where data sovereignty, latency, and resource optimization matter most.
Telecom networks generate massive telemetry and metadata every minute. This data is considered “crown jewels” by operators and must remain under strict sovereignty controls, making it unsuitable for public foundation models. Edge AI enables local processing, protecting sensitive network data while enabling faster, context-rich decisions.
Smaller, targeted models at the edge can also tackle scale better than centralized “brute force” approaches. By breaking problems into smaller, parallelizable components, operators can process tens of thousands of daily events without bottlenecks.
Observability: The Bedrock of Autonomy
Automation in telecom depends on accurate situational awareness. Yet hybrid cloud and multi-vendor networks often suffer from visibility gaps. Hickey stresses that observability across domains, vendors, and generations of equipment must be built into AI architecture from the outset. Without it, autonomous systems risk becoming ineffective or costly to re-engineer.
The trade-off is speed versus design rigor. Moving too slowly risks missing innovation windows; moving too fast without solid architecture can result in expensive technical debt.
Left Brain + Right Brain AI
GenAI excels at creative generation—drafting remediation steps, suggesting operational actions—but it struggles with network telemetry and time-series data because LLMs do not inherently understand time.
IBM’s approach is to combine “right-brain” creativity (LLMs) with “left-brain” analytical precision using Granite Time Series Foundation Models (TSFMs). These smaller, high-efficacy models are purpose-built for network data, delivering better signal-to-noise ratios and feeding insights into LLMs to produce actionable, context-specific plans.
The Enterprise and Quantum Opportunity
Hickey sees AI’s enterprise role extending beyond operations to competitive advantage —faster troubleshooting, adaptive service assurance, and optimized resource use. Quantum computing adds another dimension: it promises future leaps in processing capability for AI while benefiting from AI-driven development tools like IBM’s Qiskit Code Assistant, which accelerates quantum algorithm creation.
The path to fully autonomous telecom systems will combine observability, edge inference, hybrid cloud agility, and dual-mode AI architecture. For CSPs, this means investing in both analytical and generative AI, ensuring data control, and designing for adaptability.
As Hickey puts it, IBM’s trajectory is clear: from LLMs to TSFMs, and onward to quantum-enabled optimization—pushing the boundaries of what’s possible in intelligent networks.
Explore the full interview for deeper insights into AI’s role in telecom and IBM’s vision for the future of networking:
Read the Full Interview
#TechnicalBlog