In the pursuit of next-generation telecom operations, the industry is moving beyond traditional automation towards true autonomy, where networks don’t just follow pre-programmed rules but can sense, think, and act on their own. While intelligent systems leveraging forms of AI have been around for a while, they fall short in delivering the cognitive reasoning required for autonomous decision-making and action execution. Achieving the TMForum defined Level 4+ autonomy requires more than just smarter algorithms; it demands a shift towards systems that are goal-oriented, make context-aware decisions, and generate actions so that humans can take their hands off the operations steering wheel and play a supervisory role, and this is where Agentic AI comes in. 

What is Agentic AI?

Large Language Models (LLMs) are reshaping the way we interact with technology, enabling more natural, conversational networks and powering copilots that assist users effectively. However, they remain largely reactive: their decision-making is limited, their knowledge is bound to training data and embedded sources and is memoryless, and responses are confined to the context of the given prompt or session.

Agentic AI marks a paradigm shift toward multi-agent ecosystems capable of collaborative reasoning, leveraging persistent memory and autonomous decision-making. An LLM powered autonomous agent system is comprised of an LLM functioning as the brain, and other crucial components for planning, memory, and the use of external tools.

Self-reflection is another vital aspect that allows autonomous agents to improve iteratively by refining past action decisions and correcting previous mistakes.

• Planning helps to break down large and complex tasks into smaller, manageable steps 
• Memory helps LLM agents to learn between context in real-time and recall information over extended timeframes in the short term or long term to track trends
• LLMs leverage tools to call external resources like APIs and knowledge databases for additional information to help in dynamic decision-making. This allows the LLM to focus on tasks that it’s best suited to and use other resources for complementary information. They also use tools to perform actions, turning the intelligence into outcomes towards goal realisation.

Redefining telco operations with Agentic AI

To get the most out of AI, it must be tailored with telecom-specific domain expertise and knowledge (Why telcos must ‘verticalize’ Gen AI at scale | Nokia). With decades of telecom experience, Nokia is uniquely positioned with deep domain insight to help operators successfully verticalize AI at scale, and this is key when it comes to implementing agentic AI in operations, or AgenticOps.

AgenticOps adds a cognitive layer that bridges detection and action, where agents operate in structured workflows, applying contextual reasoning, synthesizing solutions, and implementing them using appropriate tools. They follow an adaptive execution approach, monitoring actions for their effectiveness and adjusting approaches based on real-time feedback with continuous learning. Just as specialized professionals address specific challenges, for example, doctors in different disciplines treat specific problems, specialized agents can excel in targeted tasks, while swarms of agents coordinate to achieve broader goals. This division of responsibility reduces complexity, enhances traceability, and elevates the effectiveness of AI-driven operations. 

Telco agents can make decisions aligned with strategic business objectives, whether that’s improving SLA adherence, accelerating time-to-resolution, or optimizing energy efficiency. The result is a step-change in business value, meaning fewer outages, faster service activation, better customer satisfaction, and more scalable operations, even in increasingly complex network environments. 

Agentic AI in telco operations is already shaping the future across several areas in the service operations space, like service orchestration, assurance, and security, to name a few. A few powerful examples include:

• Orchestration Agents – autonomously translate business intent into actionable workflows, decomposing tasks across multivendor domains and executing them seamlessly. This accelerates service delivery while reducing operational complexity.
• Assurance Agents – move beyond proactive detection to providing anomaly reasoning, real-time recommendations, and generating remediation steps, enabling the path to truly self-healing networks and delivering enhanced customer experience.
• Security Agents – continuously analyze threat intelligence and network telemetry to detect risks early, generate adaptive detection rules, and guide rapid remediation. The result: reduced threat dwell time and stronger network resilience.

And this isn’t just something we’re talking about. We have been putting this into practice in our recent TMForum catalyst project (Full network autonomy through intent management, enhanced observability, knowledge graphs, and AIOps-driven closed loops) and continue to evolve our approach in our upcoming project to be showcased at Innovate Asia, Autonomy accelerated: Intent to impact - Phase II. 

Bringing Agentic AI to life

As networks become increasingly dynamic and multi-layered, operational challenges will only intensify. Agentic AI has the power to make operations intelligent and effortless, enabling CSPs to shift their focus from wrestling with the mechanics of running a network to maximizing their ability to deliver customer value. But how can CSPs implement this to operate at scale whilst mitigating potential risks? What is the best approach to ensure they get the most out of this technological shift, whilst ensuring they maintain their day-to-day operations and continue to meet customer needs? Part 2 of this blog series will answer all these questions and more. In the mean time, if you’d like more information on our Digital Operations Solutions, you can find it here.