The telecommunications industry has demonstrated over the past decade how cloud-native architectures modernize traditional IT, and over the last two years validated the value of generative AI—deploying semantic search, building human-assisted tools, testing intent-based chatbots. Now, this initial phase is drawing to a close.
The next frontier is moving beyond applying AI tools in isolation at the edge of business, embedding autonomous intelligence directly into core systems. Agentic systems—networks of AI agents capable of independent reasoning, planning, and executing multi-step workflows—promise to make operational complexity transparent. Operators that deploy AI across the full technology stack—from cloud infrastructure to consumer interfaces—will define connectivity for the next decade.
Defining the Agentic Shift
Automation in the past was rigid scripting: predetermined inputs triggered predictable outputs. For example, a traditional embedded event manager (EEM) script might automatically reroute traffic when utilization on a specific router interface exceeded 90%. But if the traffic surge was caused by a software bug or cascading hardware failure, the static script could trigger blindly, exacerbating network congestion.
Generative AI ingests and summarizes large volumes of data to provide actionable recommendations, but key execution steps still require human operators. Agentic AI breaks this bottleneck, shifting the human role from manual execution to high-level governance. Powered by robust foundational models capable of parsing text, network telemetry, voice, and video, these systems not only flag issues but orchestrate the remediation process. Within safety guardrails set by humans, agents can autonomously invoke APIs to resolve network issues in real time, escalating only the most complex novel anomalies to engineers with a complete diagnostic summary.
To safely execute remediation, agents need digital twins—high-fidelity virtual replicas of the network. By combining autonomous intelligence with virtual models, operators create a safe testing environment where AI agents can continuously conduct red-team exercises and improve incident response strategies. This allows operators to test patches and configuration changes in the virtual world before deploying them to the physical network, ensuring service reliability and preventing spontaneous network outages.
This shift also reshapes customer experience. The telecommunications industry has historically had low Net Promoter Scores. Agentic AI helps the network resolve potential problems before users are even aware of an outage, improving connectivity and proactively notifying users.
Orchestrating the Agentic Digital Core
Achieving this level of operational maturity requires a structured approach that lets agentic AI directly drive the next phase of network operations evolution, targeting TM Forum maturity levels 4 and 5.
Truly autonomous networks have stalled because traditional software cannot handle unforeseen anomalies. Agentic AI enables operators to move from large monolithic applications to coordinated multi-agent ecosystems. This requires building networks composed of hyperspecialized micro-agents (e.g., billing agents, inventory systems, radio access network guardians) that communicate dynamically through standardized orchestration protocols.
This architectural loop unifies three phases of operational intelligence.First, the ecosystem breaks down data silos by unifying data platforms to perceive the environment. When intelligent agents analyze both operational network telemetry and business application data simultaneously, they can instantly understand the business impact of technical events. For instance, once a fiber optic cable is cut, the agents not only detect the hardware failure but also identify which high-value enterprise SLAs are under threat.
Second, the ecosystem reasons through intent-driven automation. Engineers no longer write heavy scripts; instead, they declare high-level business intents to the orchestration layer. For example, an engineer can instruct, "Prioritize low-latency slices for hospital remote surgery links, regardless of local hardware anomalies." The multi-agent system receives the intent, evaluates the real-time environment in a network digital twin, and coordinates sub-agents to meet business KPIs.
Finally, the ecosystem takes action by embedding self-healing loops directly into network configurations, significantly reducing mean time to repair (MTTR). Complex infrastructure anomalies that previously required an entire weekend to handle can now be autonomously corrected within minutes.
The ultimate safeguard for true agentic telecom is a deterministic governance framework. Operators must establish clear decision boundaries, allowing agents to operate autonomously within strict tolerances, but seamlessly handing over to human engineers when scenarios exceed safety thresholds. By designing safeguards directly into the core, operators can achieve machine-level efficiency without introducing systemic risk.
Return on Investment for Autonomous Operations
The industry is not hesitating to transform. Reports from GSMA and Radcom show that 71% of operators plan to deploy AI agents this year, while Google Cloud research indicates that 56% of telecom executives are already using agents in production.
The financial and operational rationale is clear. McKinsey research indicates that operators adopting AI applications for problem management have achieved a 30–70% reduction in fault tickets and a 55–90% reduction in network operations center costs.
Leading operators have turned predictions into competitive advantages:
Deutsche Telekom's RAN Guardian platform identified 237,000 network events in early 2026, reducing major network event management time from hours to approximately 60 seconds.
Bell Canada's AI operations platform leverages AI and machine learning to identify and prioritize anomalies before escalation, achieving a 25% reduction in customer-reported issues and faster MTTR.
Vodafone deploys AI agents to proactively resolve outages and optimize infrastructure expansion, protecting millions of dollars in annual operating expenditure.
KDDI has successfully deployed on-premise AI solutions in Japan’s strictly regulated market.
Agentic AI is not just a tool—it is a strategic opportunity for telecom operators to define the next generation of connectivity.