- C Spire spent eight weeks building an agentic AI system for its NOC with AWS, Chief Network Officer Alan Jones said
- Its system comprises three task agents and a supervisor agent to help find and diagnose issues
- Future agents could focus on making the system more proactive rather than reactive
Operators across the globe are trying to answer one big question: how can we use AI to really make a difference in our operations? C Spire has an answer.
The regional U.S. operator built an agentic AI system with AWS to help its network operations center (NOC) team detect and diagnose outages faster, Chief Network Officer Alan Jones told Fierce. And the early results of its efforts speak for themselves: up to an 80% reduction in mean time to detect issues, a 50% reduction in mean time to repair for non-physical issues like software tweaks and an 83% reduction in mean time to know (aka, diagnose the problem).
Stats like that are the envy of all operators. But getting there was a process, Jones said.
Building trust in the process
The key to success, Jones said, has been securing buy-in from the NOC team by including them in the agent building process.
In C Spire’s case, the new agentic system was its NOC team’s first introduction to AI tooling, Jones said. Rather than arbitrarily buying a tool and asking the team to use it, C Spire sat its NOC team down with AWS to talk through their needs. From there, C Spire and AWS built out custom agents to streamline particular tasks.
“Instead of saying ‘hey I’ve got this new AI agent that’s going to come in and it’s going to make you feel like it’s replacing you,’ it’s ‘I’ve got this new system that we’re building with AI together and I want to know what pain points, what slows you down, what keeps you from being more efficient so you can success in your job faster,’” Jones explained.
There are currently four primary agents in C Spire’s system: a knowledge base agent trained on the company’s network and documentation; a weather agent capable of checking whether network issues may correlate to storms or other weather events; an alarm agent that goes over all the telemetry data C Spire has running in AWS’ cloud; and a supervisor agent which provides a natural language interface for the NOC team to make queries and find issues faster.
All of this, Jones said, was built in about eight weeks. Much of that time was spent working with C Spire’s CISO and ensuring that the new system was built in containers no one else could access. While C Spire made certain network telemetry and alarming data accessible, it opted not to include any customer information.
“There was a lot of back and forth, a lot of architecture planning, a lot of security planning to ensure that what we were building was safe,” he said.
What’s next?
Jones said it’s possible C Spire could add more agents to the system, focused on performance and KPI monitoring. This, he added, would flip the switch from the system being reactive to proactive. But as with any AI application, doing so requires careful consideration of the data going in.
“Do you get to a point that you’ve fed so much data for proactive stuff that you’re really just causing false alarms? How do you walk that edge to make sure you don’t do that?” he said of the thought process.
Looking even further out, Jones said agentic AI could be incorporated into the middle of provisioning systems to ensure that everything runs smoothly as new services are turned up.
C Spire isn’t the only one thinking this way, either. At MWC Barcelona in March, NTT Docomo and NEC demonstrated an agentic system capable of building a complete 5G core. Nokia and Amazon, meanwhile, showcased network slicing powered by agentic AI.
Progress has been slowed by operators’ reluctance to move all of their data from on-prem environments to the cloud – for security, control and economic reasons. While Jones said trust in the cloud is increasing, it’s that pesky money bit that remains a barrier.
“I think the cloud is not as scary as it used to be,” he concluded. “We would get comfortable there if the economics worked for us.”