An article by Sasa Nijemcevic, Vice President & General Manager, IP Network Automation Business Unit at Nokia.

Nokia and other telco vendors have long advocated for the benefits of network automation. TM Forum (TMF) is also a believer, and has attempted to standardize network autonomy by establishing a structured framework for autonomous networks. However, the most compelling argument for network automation comes from the telco operators themselves.

According to a recent IDC report, improving network automation through simplification and programmability will be at the heart of European telcos’ network transformation initiatives over the next two years. This report illustrates that network automation is their highest priority, with the vast majority of telcos seeking to achieve level 4 (L4) of TMF’s autonomous networks maturity model in the medium term.

Network automation will help telcos achieve this goal in two ways. First, it will make it easier for them to reduce spending and manage the complexity of new and emerging technologies. Second, but equally important, it will empower them to take advantage of the programmable platforms they are building to grow revenues. Automation will play a key role in addressing these challenges, particularly with technologies such as network slicing. It will allow operators to use their resources efficiently and deliver services to their customers with quality and in a timely manner.

A scalable and trustable automation platform

Telcos can’t build an automation framework by just throwing tools and scripts at the problem. They need a thoughtful and structured approach that addresses the fundamental elements of automation, including standardization, modularity and integration.

TMF’s autonomous network framework provides a useful baseline for evaluating network autonomy. The autonomy levels are evaluated based on task ownership in the network’s full management cycle. As more tasks that require cognitive abilities are performed by systems as opposed to humans, the higher network autonomy levels are achieved.

This is a useful guideline for network operators taking steps toward higher network autonomy. It will help them build a scalable automation platform with the right elements in place to speed up progress toward their goals. Let’s look at some of the fundamental elements for this journey.

Intent-based networking

Trusting the system to define and fulfill intents for network services and achieve L5 network autonomy as outlined by TMF may seem like a future goal. However, operators can attain L4 through a programmable and scalable automation platform. Such a platform allows operators to plan and describe the intended network services and let the system take care of the implementation details, abstracting network complexity and vendor variety.

The intent-based approach ensures that network configurations are consistent and enables model adaptations to meet the demands of new service offerings in a timely and cost-effective manner. While it has numerous advantages, the intent-based approach requires a shift for service providers. It moves the source of truth from the network to the intent.

Service assurance for intent-based networks

Defining and fulfilling intents for network services is not limited to configuration. It also involves assuring that the service will deliver the expected quality and often rerouting the traffic to meet service-level agreements (SLAs) with customers. Many automation platforms allow for both but as multiple separate operation streams, which can bring about complexities and be difficult to align. Intent-based service assurance that combines service health into a health indicator can help operators avoid this complexity by monitoring and assuring the service as described by the network planner and triggering an action before a service outage occurs or SLAs are compromised.

Generative AI assistant: A natural complement

According to a Nokia Bell Labs forecast, artificial intelligence (AI) traffic is expected to account for 33% of global traffic by 2030. This includes traffic generated by new AI-powered applications such as ChatGPT and indirect traffic generated as a result of AI algorithms influencing and increasing user engagement. These applications require stronger network performance, including lower latency, higher availability and enhanced security. They also need networks to evolve and improve accordingly.

Traditional AI and machine learning have already been used to augment the acuity of operations staff by providing capabilities such as baselining, outlier detection, capacity planning and traffic prediction. The advent of generative AI (GenAI) is now taking this to the next level by augmenting their expertise through natural-language AI assistance in areas such as:

• Documentation: Providing explanations and guidance on network functionalities, which reduces the entry barrier for less-skilled operators.
• Troubleshooting: Offering real-time network health and performance reporting and recommending next best actions for issue resolution.

• Coding and programming: Generating software artifacts such as intents and workflows.

Agentic AI: Futuristic or unfolding?

An AI agent has been defined as a software entity that perceives its environment, makes decisions and acts autonomously to achieve specific goals. This concept may seem to come from a science fiction movie such as The Matrix or a time far in the future. Based on this definition, however, it’s easy to identify non-AI agents in our ecosystem today. And it is not unimaginable that we could augment their capabilities by training them with appropriate network data and enabling them to take actions. A programmable automation platform can collect vast amounts of data from a heterogeneous network, unify it and enable closed-loop actions—essentially enabling AI agents.

Consider the Path Computation Engine (PCE) as an example. A PCE monitors the network and, based on predefined policies (e.g., preferences set by network engineer) or telemetry collected in real time (e.g., detecting congestion), decides to reroute the traffic using closed-loop action capabilities. If an AI agent is trained using the telemetry data collected from the network, the history of past actions and their outcomes, the policies and preferences defined by the network engineer and any other relevant data, it will be able to optimize network traffic autonomously.

Conclusion

The necessity of network automation has been well established, and many operators are making progress on their journey toward building highly autonomous networks. However, the requirements for automation platforms are changing rapidly, driven by the demands of data centers acting as the fabrics for AI applications. These demands are imposed on the networks that are connecting them, especially those with mission-critical priorities.

At Nokia, we are committed to providing our customers with a flexible and agile network automation platform that will enable them to manage increasingly challenging network requirements. Our platform helps them take on every demand, whether it’s delivering services with expected quality, slicing the network, or detecting and resolving network problems in real time to deliver the requested bandwidth and uphold strict SLAs.

Looking to the future, Nokia Network Services Platform (NSP) is ready to scale and adapt to enable agentic operations and leverage GenAI as a complement to network automation. This evolution will enable service providers to build autonomous networks that can sense, think and act!