Service assurance is officially graduating from an era of dashboards, tickets, and engineers scrambling to find what’s gone wrong to swift root cause analysis and proactive fixes
As AI moves deeper into the network stack, a burst of experimentation has followed to figure out how to best tune the network with AI.
“The networks today are 150x more complex than legacy networks and the only way to address or manage this operational complexity is through continuous testing and total automation,” noted Anil Kollipara, VP of product management at Spirent in the recent presentation.
Over the past few months, a clear trend has emerged: solutions providers are embedding AI into their portfolios to unlock greater levels of autonomy, observability, and speed of resolution. The goal is to make service assurance low-touch for operators, for many of whom complete automation of service assurance processes remains a near-term goal.
This change was long in the coming. Network operations has had an ill reputation for quite some time. It’s viewed by insiders as a thankless job, involving long shifts, tedious tasks, and finger-pointing when things go wrong.
Now as the responsibility of network testing and service assurance has shifted hands from equipment vendors to service providers, there is a natural urgency to figure out how to improve service quality controls and cut repair time.
There is evidence that points to the fact that the degree of autonomy in service assurance has been on the rise among operators. A GSMA Intelligence report finds that three-quarters of the operators surveyed are in the process of automating their service assurance processes, while over a third indicated that a majority of their processes are already automated.
Although AI may not take all the credit yet, but AI-driven service assurance is definitely gaining steam among operators. Crucially in three areas, AI’s role is becoming increasingly vital across domains.
Root cause analysis
“The process of getting to the bottom of a problem, the whole root cause analysis (RCA), is a very painstaking and tedious process even with an automation cycle put in place,” observed Kollipara.
There are multiple steps to RCA, including but not limited to defining the problem, gathering artifacts, running analysis, making diagnosis, and identifying the root cause
— that makes it trying.
AI offers some very specific capabilities that cut this weeks-long process to minutes. For example, it can scan through large volumes of datasets almost instantly, identify patterns in them, and make automated correlations across systems.
That makes connecting the dots which is essentially the root cause analysis exercise a lot easier and reliably automatic. Within minutes, AI can look through thousands of data points from network logs, telemetry and KPIs and reveal where an incident occurred and what caused it.
Currently, according to some research, RCA is one of the top AI use cases in telco networks.
Proactive anomaly detection
AI workloads are chaotic, in lack of a better word, which invites frequent anomalies and deviations.
AI models present an unique opportunity to resolve them. Good AI models can spot unusual patterns or outliers in large datasets with 100% accuracy, and that’s a great way to catch performance deviations in networks.
As AI continues to make networks wildly complex, on the reverse side, it is helping providers cut through that noise and proactively detect issues ensuring fewer outages.
With level-4 and level-5 autonomy being the ambition for most operators, AI-driven proactive anomaly detection is believed to be one of the fastest ways to get there.
Customer analytics
AI-driven analytics is another one of the most practical AI use cases in service assurance. AI models are good at reading user experience degradations, usage patterns, upselling, and other analytics, that can indicate churn. This allows them to foresee risks of customer loss and
The GSMA report finds that a majority of operators already use AI for customer analytics, with 80% using it to generate customer-related insights, and 63% for customer complaint analysis. An additional 34% indicated that 51% to 75% of their analytics processes today are AI-driven.