Operators must move beyond basic human-in-the-loop assistance to embed agentic AI across their digital core systems, build invisible self-healing networks, unlock billions in value, and define next-generation connectivity – writes Google Cloud.
The telecommunications industry spent the last decade proving that cloud-native architecture could modernize legacy IT. It spent the last two years proving the value of generative AI – deploying semantic search to map vast knowledge bases, building human-in-the-loop assistants to boost front-line staff efficiency, and testing early intent-based chatbots. That introductory chapter is closing.
The next frontier for telecom operators is to move past bolting isolated AI tools onto the fringes of their business and start integrating autonomous intelligence directly into the core. Agentic systems – networks of AI agents that can reason, plan, and execute multi-step workflows independently – have the power to make operational complexity invisible. The carriers that deploy AI across their full tech stack, from cloud infrastructure to consumer-facing surfaces, will define the next decade of connectivity.
To understand why agentic AI is poised to reshape telecom requires a look at what separates this generation of technology from everything that came before it.
Defining the agentic shift
Legacy automation from the last decade was rigidly scripted: a predetermined input triggered a predictable output. For example, a traditional Embedded Event Manager (EEM) script might be programmed to automatically reroute traffic if a specific router interface exceeded 90% utilization. But if that traffic spike was caused by a sudden software bug or a cascading hardware failure three nodes away, the static script would blindly fire anyway, often worsening the network congestion.
Generative AI solutions improved on legacy systems by ingesting and summarizing massive amounts of data to offer actionable recommendations. Yet, the critical final step – actual execution – still required a human operator to manually implement the fix.
Agentic AI breaks this bottleneck by shifting the human role from manual execution to high-level governance. Powered by advanced foundational models capable of parsing text, network telemetry, speech, and video across massive context windows, these systems don’t just flag problems; they orchestrate the remediation. Working safely within human-defined guardrails, agents can autonomously trigger APIs to resolve network issues on the fly. Instead of forcing a human to run the playbook, the system executes the fix independently, escalating only the most complex, novel anomalies to engineers with a complete diagnostic summary.
To execute these fixes safely, agents require digital twins – high-fidelity virtual replicas of a telecom network. By pairing autonomous intelligence with these virtual models, operators create a safe-to-fail environment where AI agents can continuously run autonomous red-teaming exercises and enhance incident response strategies. Crucially, this setup allows operators to test network patches and configuration changes in the virtual world before deploying them to the physical network. This helps operators guarantee ironclad service reliability and prevent the accidental, self-inflicted outages that frustrate subscribers and erode brand loyalty.
This shift also rewrites the narrative on customer experience. The telecom sector has historically suffered from low Net Promoter Scores. Agentic AI helps the network resolve a brewing localized issue, improve the user’s connection, and quietly notify them before they ever realize a disruption occurred.
Orchestrating the agentic digital core
Transitioning to this level of operational maturity requires a structure where agentic AI directly enables the next evolution of network operations, aiming for TM Forum Level 4 and 5 maturity.
True autonomous networking has historically stalled because legacy software couldn’t handle unpredicted operational anomalies. Agentic AI moves operators away from massive, monolithic software applications toward coordinated multi-agent ecosystems. This requires operators to build a fabric of hyper-specialized micro-agents – such as billing agents, inventory systems, and radio access network (RAN) guardians – that communicate dynamically using standardized orchestration protocols.
This architectural loop unifies three distinct phases of operational intelligence.
First, the ecosystem perceives the environment by breaking down data silos via unified data platforms. When agents can reason across operational network telemetry and commercial application data simultaneously, they instantly understand the real-world business impact of a technical event. For example, if a fiber optic line is cut, the agent doesn’t just see a hardware fault; it instantly recognizes which high-value enterprise SLAs (Service Level Agreements) are threatened.
Second, the ecosystem reasons through intent-driven automation. Instead of forcing engineers to write prescriptive, code-heavy scripts to manage traffic spikes, they simply state a high-level business objective – or “intent” – to the orchestration layer. For instance, an engineer can issue a directive to “prioritize and preserve the low-latency slice for hospital remote-surgery links, regardless of localized hardware anomalies.” The multi-agent system receives this intent, evaluates the live environment on a network digital twin, and coordinates which sub-agents must act to satisfy the business KPI.
Finally, the ecosystem takes action by embedding self-healing loops directly into network configurations, drastically slashing Mean Time to Repair (MTTR). Complex infrastructure anomalies that used to trigger weekend-long engineering crises are transformed into minute-long, fully autonomous corrections.
The final guardrail of a true agentic telecom is a deterministic governance framework. Operators must establish explicit decision boundaries where agents operate autonomously within tight tolerances, but seamlessly hand off to human engineers the moment a scenario crosses safety thresholds. By designing these guardrails directly into the core, operators can capture machine-speed efficiency without introducing systemic risk to critical infrastructure.
Proving the ROI of autonomous operations
The industry is not hesitating to embrace this shift. A recent report from GSMA and Radcom found that 71% of operators plan to deploy AI agents this year, while a Google Cloud study revealed that 56% of telecom executives already have agents in production.
The financial and operational justifications are clear. Research from McKinsey & Company indicates that operators using AI applications across the issues-management journey achieved a 30 to 70% reduction in troubleshooting tickets and a 55 to 90% reduction in network operations center (NOC) costs.
Forward-thinking carriers are already turning these projections into competitive advantages:
- Deutsche Telekom’s RAN Guardian platform identified 237,000 network events in early 2026, shrinking the time required to manage major network incidents from hours to roughly 60 seconds.
- Bell Canada’s AI Ops platform uses AI and machine learning to identify and prioritize anomalies before they escalate, achieving a 25% reduction in customer-reported issues and a significantly faster MTTR.
- Vodafone deployed AI agents to proactively resolve outages and optimize infrastructure scaling, protecting millions in annual operational expenditures.
- KDDI, in a highly regulated Japanese market, successfully deployed on-premise AI tools that enable fully automated network operations while strictly adhering to sovereign data laws.
Reclaiming the innovation mandate
Ultimately, the agentic era will be judged by financial performance. Appledore Research estimates that agentic AI could drive $60 billion in operational cost savings for the telecom industry by 2030, while Deloitte predicts that the technology will unlock $150 billion in total value. It is no surprise, then, that Google Cloud found that 55% of telecom executives are allocating 50% or more of their future AI budgets to AI agents.
The carriers moving first are redefining what operational maturity looks like. They aren’t adopting AI simply to write better marketing copy or power basic chatbots. They are deploying autonomous agents to make the underlying network invisible, resilient, and effortlessly efficient. In 2026, the infrastructure to deliver on that promise is finally here.
Anil Jain is the Global Managing Director for the Strategic Industries at Google Cloud, with responsibilities spanning Media & Entertainment, Telco, Gaming, Retail, Consumer Packaged Goods, Financial Services, Healthcare & Lifesciences, and a number of Industrial industries. A veteran of the technology industry, Jain is focused on helping organizations digitally transform the way they run their businesses through the application of AI, data and cloud-based innovation to deliver new consumer experiences.
