Luma seeks to modernize traditional approaches to root cause analysis with a powerful combination of agentic and GenAI
The rapid advances in agentic AI has stirred strong interest among network and service assurance providers, many of whom, mired in slow, exhaustive grunt work, are now seeing a way out.
Spirent is the newest member to join a fast-growing community of providers who have embraced agentic AI. The company just released its first agentic AI solution, Luma, which is designed to analyze network issues at top speed.
Agentic AI boasts of an extraordinary degree of independent execution, and when combined with contextual awareness and intelligence, it becomes a valuable tool for triaging issues.
“Luma applies both to our lab products and our assurance products that we’re using for use cases like root cause analysis,” said Ross Cassan, senior director of assurance strategy at Spirent.
Available via their signature testing platform, Landslide, Luma takes on a weighty challenge in network testing. “One of the issues our [customers] have to deal with is once something starts happening in the network, determining what the blast radius of that problem is, is something the management team wants to know.”
That seemingly simple reporting task has to go through a winding process of deep-diving into the issue, tracking its origin, and mapping its blast radius. For engineers deep into troubleshooting work, that hunting and pecking exercise is like trying to fix the plane while flying it.
“They get ripped in two directions. And that problem of reporting to the management team what the blast radius of the problem is, is a great use case for AI,” Cassan said.
Luma addresses the issue in two ways: the multi-agent solution, trained on the vendor’s network knowledge base, sits directly into the customers’ testing workflows. This allows it to access and analyze test results more pointedly.
It can easily trace where a problem is, what devices are connected with it, how far the outage spreads, etc. pin-pointedly. “We focus on being able to provide visibility of all of them,” Cassan added.
Luma rises above the league of passive alerting tools and serves as a source of rich network intelligence. Cassan emphasized that the agents underpinning Luma are trained on domain knowledge.
“Our systems have always been plugged into the network database of nodes. So we know what the architecture of the network is and what the topology looks like. One of the things we don’t want to do is hit the alarm when a system is offline for maintenance for example. So we develop that knowledge and bring that together with our knowledge of testing, our knowledge of 3GPP specs and how different nodes behave is really the key to our agents and making sure that they have the guardrails so that they’re not just relying on general knowledge to make their claims.”
Additionally, it comes with a conversational interface which is useful for querying information in natural language.
“Having the chatbot helps our engineering users to understand what’s happening and really hone in on the problem….and the customer can ask it about different circuits, different network elements, different regions, different services, and ask for performance graphs,” he explained.
Amid the brain drain happening across tech, Cassan hopes the solution will help support the existing crew by amplifying their output.
“We’ve seen a reduction in the number of subject matter experts in our customer base and being able to supplement them and help out the people that are there is kind of where we found the key problem. So, that’s where we’re applying most of our focus,” he said.
Ultimately, the chatbot’s ability to summarize issues succinctly in plain English language makes the reporting job that much easier for engineers.
