We're in the middle of a huge fiber broadband roll out. As more things and more people connect - and demand higher performance - operating a network is becoming increasingly challenging. By now, we have passed the point where humans alone can understand the full complexity of the network. Operators must find ways to meet and exceed end-user’ expectations and be successful in competing for a finite pool of FTTH customers. In this environment, data-driven analytics and AI/ML are becoming a vital tool for network operators.

Improving data collection and service assurance can be challenging when relying on the traditional semi-automatic processes. AI-driven operations (AIOps) can compress the time it takes to evaluate network health, generate new insights from data, and solve network issues in a proactive way. According to a 2024 survey by Nvidia on the state of AI in telecommunications, customer experience (49 percent), Generative AI (42 percent), predictive care (37 percent) and planning & operations (34 percent) are the key areas operators are investing in.

Networks that sense, think and act

AIOps fits within a network that senses, thinks, acts. AI, of course, does the thinking. But first it must sense what is happening in the network, and this is done through modern network data telemetry that collects, processes, and stores vast arrays of network data for AI to work with. Real-time analysis can accurately identify patterns in data that represent issues, which makes AI highly reliable to report incidents. By extension, it can also predict future trends and their probability of occurring, allowing human operators to take preventive action to reduce outages.

There are four broad areas where AI technology really makes a difference: in providing assistance to humans (yes, we still need humans), in analyzing data, in driving digital twin analysis, and in supporting autonomous network operations. This blog is going to touch on AI’s role in all of these and will be followed by additional posts diving into specific examples and use cases.

Assistance to humans

A great example of an AI-assisted human activity is the use of “computer vision” in field work. A technician installing an access node, for example, takes a few photos of the installation and uploads them. AI image recognition can verify the installation, connections and labelling, instantly update the network inventory, and correlate service alarms with both passive and active network inventory. Another example is one we’re all familiar with: natural language processing in a chatbot-like interface (think ChatGPT) can search, filter, summarize, and explain network status, and even generate dashboards and reports, or suggest actions to solve a network issue. 

Anomaly detection

An extremely powerful use case for operators is AIOps’s ability to crunch vast amounts of data and detect network anomalies. An anomaly is a pattern in the data that differs significantly from what is considered normal or expected behavior. AI can easily outperform static network-wide rules, which require domain knowledge and manual effort to maintain. In addition, AI can deal better with increasing data volumes and even anticipate issues based on emerging trends or by comparing behavior between neighboring data sets.

Digital twin

The digital twin concept is brilliant for network management. Load all your sensed network data into a digital twin of your network and run AI-based models, simulations or scenarios to see what happens. If all goes to plan, you then apply your model to the live network. Capacity planning, bandwidth management and network traffic modeling can all benefit, helping operators meet and enhance SLAs, plan corrective steps, or know where to market new tiers of service.

Autonomous operations

While anomaly detection smartens the sensing and the digital twin network supports the thinking, autonomous operations enable to take actions. Autonomous networks can make decisions and self-optimize locally based on predictive analysis or closed-loop automation. For example, by monitoring the state of a network service or asset, AI can determine the best possible action to restore the health and correct an undesired state. Based on an operator survey on TMForum’s Autonomous Networks, Capgemini found that organizations with an advanced implementation strategy in Autonomous Networks outperform organizations without – lowering network opex by nine percent in the next two to five years.

We’re already seeing the value of automation and AIOps in broadband networks and that value is increasing, fast: in fact, the progress in AI is developing more rapidly than probably any technology in recent memory. If you look at Gartner’s last Hype Cycle for Artificial Intelligence, you can see AI applications reaching peak productivity in the coming years. Broadband operators are primed to reap the benefits.

Watch out for the next blog in this series and, in the meantime, download this research paper to learn more. Also, why not check out the Nokia Altiplano Access Controller, the domain controller that brings sophisticated intelligence into the fixed access domain, which was awarded the #1 network automation platform for broadband access networks worldwide in 2024.