When people ask me what an autonomous network is and how it differs from a conventional automated network, I usually start with a simple distinction: automation is about the ability to execute, while autonomy is about the ability to make decisions.
An automated network simply follows fixed rules and reacts to predefined events. An autonomous network, by contrast, is aware of its own state, its objectives (intents), and its environment.
Because automated systems operate according to fixed rules, those rules constrain what the systems can actually do. Today’s networks have become so complex that managing them through large sets of predefined rules is increasingly impractical. This is why autonomous models are emerging that can self-regulate without relying on unmanageable sets of rules and static parameters predefined by humans.
Why Now?
Complexity is one of the main reasons why the concept of autonomous networks is gaining so much momentum among major operators today. They make the technical and operational complexity associated with 5G and IoT manageable; otherwise, certain aspects of these technologies would be impossible to manage. At the same time, the industry needs to expand its service portfolio while improving operational efficiency. The Autonomous Networks framework is designed to support both objectives at once.
Who Benefits?
At the highest level, the entire ecosystem benefits: users, operators, and new ecosystem partners. At first glance, it may seem that autonomous networks primarily benefit the largest operators, but I do not believe the opportunity is limited to them. Small and medium-sized ISPs also need to assess both the potential benefits of embracing autonomy and the risks of falling behind. It may seem difficult to take the first step, but at the right scale and in response to genuine needs, I believe it is imperative to get started. Once large operators have gained the capabilities that autonomous networks bring, they will hold a competitive advantage—both in terms of cost and the user experience they provide—putting medium-sized and small operators at a disadvantage.
From an Infrastructure Management Perspective
From the standpoint of infrastructure management and operations, I would first argue that TM Forum’s Autonomous Networks concept might more accurately be called the “Autonomous Telco.” The word “network” can make the framework sound as though it is limited to technical infrastructure, whereas its scope actually extends across the entire organization.
In fact, the architecture defines a three-layer model: the resource layer, which is the most technical, and the service and business layers. Therefore, adopting the Autonomous Networks concept actually means embarking on a journey to transform the company’s entire operating model.
With that in mind, starting at the resource layer and gradually expanding makes sense. Even focusing initially on a single Autonomous Domain—as defined by TM Forum—such as the access network or the IP domain can lead to better resource planning, improved availability management, and faster service provisioning and change implementation. These are tangible benefits that can emerge relatively early in the journey.
Day-to-Day Operations
The impact on day-to-day work is gradual. Before looking at how those activities change, however, it is useful to introduce two concepts that help clarify what autonomy means in practice.
The TM Forum framework introduces two key concepts. The first is a set of autonomy levels ranging from L0, where operations are entirely manual, to L5, representing full autonomy—still more of an aspirational state than an immediately achievable one. The second is the closed-loop model, commonly expressed as AADE, or I-AADE/IAADE. Without going into detail, it represents the different steps involved in the processes required for any operation: collecting information, analyzing it and suggesting a course of action, making a decision, executing it, and finally reviewing its effects. When we measure the level of autonomy, what we are essentially assessing is whether each part of that process is performed by a human or by a system.
How daily work changes therefore depends on the level of autonomy achieved. As autonomy increases, people gradually step out of specific functions within the loop and move into different roles. Instead of executing individual tasks, teams increasingly supervise, govern, and operate the processes that have been made autonomous.
The Push from Generative AI and Its Challenges
The autonomous networks concept came before generative AI. That said, generative AI, and now the notion of agentification, are giving the model a major push forward.
The technology required to transform the way we operate networks is now within reach, but significant challenges remain. Some are common to other industries deploying AI in areas such as customer service or sales. Telecommunications, however, introduces additional considerations because these systems are interacting with critical network infrastructure and service platforms supporting thousands or even millions of users.
In my opinion, during an initial stage it is advisable to work on all the stages of the closed loop described above except decision-making. At that stage, I believe the Human-in-the-Loop concept should be maintained until full confidence has been achieved in the implemented use case.
Implementation should be gradual: as the saying goes, you eat an elephant one bite at a time. A practical approach is to focus on a single Autonomous Domain—for example, a particular technology domain—and pair it with a specific operational process such as fault management. Together, these form an evaluation object under the TM Forum model. From there, each subscenario can be addressed progressively, one step at a time.
Initially, we will have hybrid processes. Moving from performing operational tasks to exclusively supervising processes will take time, but this is the path toward changing the operating model.
Getting Started: Operators and ISPs
The first step is to embrace the concept, because doing so leads to a different way of approaching problems. Depending on its organizational DNA, each company may need either a top-down or bottom-up model to drive the initiative. A large organization will probably require a cascading approach led by top management. In smaller organizations, however, I see a bottom-up approach as viable, where technical teams stop focusing solely on individual automation initiatives—which they are likely to implement anyway—and begin thinking within a new framework. The amount of work required is similar; what changes is the approach.
As described above, the next step is to select a specific evaluation object and concentrate the initial effort there. It is equally important to recognize that building autonomous capabilities is inherently multidisciplinary. This is not a matter of Engineering or IT telling Operations what to do. It requires cross-functional teams to work through problems together, bringing perspectives that may initially differ but ultimately lead to better solutions.
The first use case presents challenges from every angle, but the main challenge is learning to think differently and ensuring that everyone participates. Over time, I have seen that cultural barrier begin to fade. Once it does, teams can devote more of their attention to the underlying technology challenges.
To explore firsthand how to take this step toward autonomous networks and transform infrastructure management, we invite you to join us at LACNIC’s upcoming event in Mendoza for the panel “Autonomous Networks: Automation for More Agile Operations”. My colleagues will present real-world use cases, strategies for gradual implementation, and practical lessons for building networks that are more efficient, scalable, and ready to address the challenges facing our region. Register here. We hope to see you there!
