Hakan Ekmen is CEO of P3 communications and P3 re:invent, helping leaders turn technology into measurable performance and market advantage.
Artificial intelligence in telecommunications is no longer a future conversation. Most operators have already embedded machine learning into parts of their organizations; many are actively piloting generative AI; and a growing number are now exploring agentic systems, or AI architectures capable of reasoning across tasks and interacting with multiple data sources.
This momentum is encouraging. Yet, in boardrooms and executive forums, a practical question increasingly surfaces: How does agentic AI become truly actionable—not just innovative in prototypes, but operationally and financially valuable in live networks?
The good news is that the answer is clearer, and more achievable, than many assume.
From Models To Systems: Where Value Emerges
Much of today’s AI discussion still centers on model size, benchmark scores and stylistic fluency. These dimensions matter, but in telecommunications, success is measured differently. Performance is measured in network stability, operational efficiency, investment precision and customer experience.
The most important shift, therefore, is not from smaller models to larger models, but from isolated models to integrated systems.
Agentic AI becomes actionable when it evolves from a conversational interface into a retrieval‑augmented, decision support layer embedded directly into operational and commercial environments. Intelligence must connect seamlessly to live networks, operational workflows and business systems.
Pragmatism matters here. For repeatable and deterministic tasks, traditional rule‑based automation often remains faster, more reliable and more cost‑effective than agentic AI. A site outage caused by power loss or hardware failure can be detected efficiently using established operations support system (OSS) alarm filtering.
Where agentic AI adds distinctive value is in reasoning across complexity—correlating local faults with cascading performance degradation, service impact and customer experience signals. Combining deterministic automation with reasoning‑based intelligence lays the foundation for real operational impact.
A Practical Framework: Retrieve, Reason, Recommend
From an operator’s perspective, the value of agentic AI can be understood through three reinforcing capabilities.
1. Retrieve: Making Enterprise Knowledge Accessible
Telecom organizations possess vast and diverse data assets: network telemetry and alarms, infrastructure documentation, OSS and business support system (BSS) logs, customer interactions, billing records and marketing data.
Agentic AI becomes powerful when it retrieves the right information at the right moment across these domains using retrieval‑augmented architectures. Rather than relying only on static or pretrained knowledge, insights are dynamically grounded in current enterprise data. This grounding improves trust, traceability and explainability—essential prerequisites for operational adoption.
2. Reason: Connecting Technical And Commercial Context
Telecom challenges rarely exist in isolation. A congestion issue may involve radio constraints, usage behavior and investment timing at the same time. A spike in call center volume may stem from a network change, a billing anomaly or a marketing campaign.
Agentic AI becomes actionable when it can reason across domains by detecting multilayer patterns, identifying correlations and inconsistencies, connecting technical signals with business indicators and simulating decision scenarios before execution. This reasoning layer transforms raw data into situational awareness.
3. Recommend: Supporting Confident Decisions
The final step is delivering recommendations that are decision‑oriented and explainable.
• Should capacity in a region be expanded now or reviewed next quarter?
• Is a campaign driving profitable growth or unintended network strain?
• Can parts of the upgrade roadmap be deferred without risk?
• Is a billing anomaly isolated or systemic?
Agentic AI supports leadership decisions rather than replacing them. By improving speed, clarity and confidence, recommendations enable organizations to act earlier and more effectively.
From Proof To Impact: A NOC Example
A tier‑one operator seeking to strengthen fault and incident management faced rising priority incidents, overlapping tools and growing customer escalations. An agentic AI system was introduced to complement existing network operations center (NOC) workflows.
The system ingested real‑time alarms, logs and telemetry; correlated multilayer faults into unified incidents; assisted triage with probable root cause and impact estimation; recommended runbooks and resolver assignments; and identified patterns signaling repeat incidents.
Within three months, results were tangible: 30% to 40% reduction in mean time to repair; 20% to 25% fewer duplicate and false‑positive tickets; 15% to 20% reduction in field dispatch costs; improved SLA adherence and customer experience; and significantly reduced cognitive load for NOC teams.
The technology amplified human expertise, allowing engineers to focus on resolution quality rather than alert volume.
Actionability Where It Matters Most: OpEx And CapEx
Operational expenditure (OpEx) in telecom extends far beyond the network. Predictive maintenance and automation reduce technical OpEx. But agentic AI becomes more powerful when it connects technical and commercial domains, including marketing effectiveness, customer care volumes, billing accuracy and service management. By correlating these dimensions, leaders gain a holistic view of cost drivers and value creation.
The same applies to capital expenditure (CapEx). Instead of relying solely on forecasts and buffers, investment decisions can increasingly reflect real utilization patterns, demand signals and customer behavior. The result is smarter spending and more confident, defensible investment planning.
Architecture, Talent And Ecosystem
Telecom environments evolve continuously. Networks change. Regulations shift. Services expand.
For agentic AI to remain actionable, three enablers matter equally:
• Modular, adaptable architecture integrating legacy and cloud‑native systems
• Strong data and AI foundations with governance, security and clear guardrails
• Skilled talent combining telecom domain expertise with data and AI capabilities
Ecosystem partners also play a key role in accelerating integration and helping operators move from pilots to production. When accountability, capability and architecture align, AI becomes part of the operating model rather than a side initiative.
From Experimentation To Operating Model
Today’s telecom leaders must balance network complexity, cost efficiency, investment decisions and rising customer expectations at the same time.
Agentic AI becomes actionable when it helps manage this full picture as a decision support layer embedded into daily operations.
The real breakthrough is not artificial intelligence itself. It is making intelligence operational.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
