Generative AI could democratise network operations

Network operations is an expert field, and there's a good reason for that. These specialists comb through a network's raw data and analyse the problems and issues that emerge from it. In doing so, they keep the network—and business—running smoothly. It’s a crucial function within the modern network, yet one that is facing an ever-mounting degree of friction.

The ever-growing enterprise network

The average business network has grown hugely in recent years. This age of digital transformation has introduced well-worn legacy infrastructure to new technologies. The IoT and the cloud have all fundamentally changed the way networks operate. 

They’ve done this in two principal ways. Firstly, they've fundamentally changed the architectural form of a network. As new technologies and assets have been introduced, the network expands outside the sight or traditional range of vision, sitting outside the perimeter - such as IoT and cloud - and leaving a series of blindspots that teams cannot see into. One report from Dimensional Research estimates that 81% of operations teams struggle with network blind spots. Those blindspots create effective new breach points for attackers, new, unexpected pathways for traffic to emanate from and - to boot - new types of traffic. 

Perhaps unsurprisingly, businesses now report that these new technologies and types of traffic are exactly the places they report visibility problems within. LiveActions 2024 Network Performance and Monitoring Trends report (NPMT) showed 57% of organisations noted that they lacked visibility into relatively new areas like Cloud SD-WAN, Voice and Video. 

Secondly, these transformations exponentially increased the load of traffic on a network. Some estimate that the average organisation’s traffic volume has doubled, if not tripled, in the last five years. 

The 2024 NPMT showed that respondents' current network performance management (NPM) solutions cannot keep up. Almost half of all respondents—49 %—say that their current NPM can’t provide insights that help them solve their current network problems. However, perhaps more importantly, 43% say that their NPMs cannot scale with the data volumes and network complexity they now deal with.

Tool sprawl

While the network management landscape has grown uncontrollably, the tools they use to manage it are often legacy solutions designed to view a type of network which we’ve long innovated in the past. On top of that, these tools usually accumulate quickly; according to one study, the average NetOps team uses up to 10 tools to monitor their network. According to another survey, almost a quarter of large enterprises rely on up to 25 network performance monitoring tools. 

These are complex pieces of technology with their own user interfaces, programming languages, and metrics, and they often don’t integrate. This creates a variety of data streams that look at different aspects of network operation but little understanding of the true nature of network behaviour. This kind of tool sprawl creates an inconsistent patchwork vision of the network, which ultimately increases confusion when it is meant to eliminate it. 

Skills gaps

Dealing with this complexity is a series of over-stretched and understaffed network and security professionals. Much has been made of the global IT skills gap, with IDC predicting that by 2025, 90% of global organisations “will be staring down the barrel of a crippling IT skills gap.”

The same is true in this case, where cybersecurity and network operations skills are in high demand while supply seems stubbornly lacking.

This results not just in skill deficits around critical network functions but ultimately stretches the time, focus, and energy of the experts present. Their effectiveness as experts must improve as they chase down false positives and manoeuvre between sprawled stacks of legacy tooling. 

Generative AIs offer a way out

In short, many have arrived at an impasse. The load of increasing demands on network operations is now colliding with decreasing capacity. Yet, a new possibility exists in the explosion of generative AI technology that has ignited in the last few years. 

The key here lies in its ability to take dense technical information and translate it into natural language. That's the key that could allow organisations to tame the uncontrollable growth of their networks and diminish their capacity to mitigate and monitor it. 

This works in several ways. First, it mitigates complexity by offering natural language explanations of otherwise complex network issues. 

If, for example, an NPM tool shows packet loss in a particular part of the network, a generative AI could deconstruct the event by offering an analogy, comparing it to a traffic jam that occurs at peak driving times and rush hour. 

Secondly, one of the big problems that network engineers face is drawing a line between a tool's analysis and an actionable insight that might tell them what to do next. When an NPM tool shows that data transfers are slowing due to saturated bandwidth in the data centre, a generative AI can suggest implementing link aggregation by binding network links into a single connection, thus expanding the lane down which data transfers travel. 

Finally, it fundamentally lowers the entry barrier for these otherwise specialised roles. Network operations staff are experts in their field, and that expertise can’t be taken for granted. However, Generative AI provides a novel way to train new experts from existing staff and give precious time back to those network operations veterans whose attention and expertise are valuable. Take the example of the protocol analyser - also known as a packet sniffer - which is one of the most essential tools in the network operations toolkit. When such a tool reveals, for example, a spike in TCP retransmissions, fledgling network engineers can use a generative AI to explain what that means and how to deal with it.

These technologies have come just in time for network operations specialists. They’ve spent the last few years seeing their workloads grow and their networks complex, putting out fires while they should have been using their expertise on strategic problems which might help a network innovate and enable business growth over the long term. Generative AI can help collapse that mounting complexity and open the discipline of network operations to more people, thus expanding the talent pool a business can draw on to live up to this crucial function.