9
Operators are seeing more success with agentic AI where use is localized, not cross-domain, said Patrick Kelly, Appledore Research

Although the telecom industry has been typically slow in its uptake of new technologies, a shiny new tech has been peppering conversations these past few months. It’s agentic AI. 

Communications service providers (CSP) are getting up to speed with agentic AI, with many beginning to experiment with agentic tools to refine their operating models. 

In a newly-published report, Appledore Research digs into the early business impacts of agentic AI in telecommunications. In this episode of RCR TV program, Pulse, I sat down with Patrick Kelly, founder and principal analyst, to understand some of the ways operators are using the technology today — and what lies ahead. 

For years, rule-based systems and analytics have been CSPs’ go-to for network operation. But complexity sprawls, growing customer expectations, and intermittent network issues causing sluggish services in peak hours, are forcing them to look for new ways to optimize operations. 

During his research over the past five months, Kelly has found that the impact of agentic AI has been profound across CSPs, but still “tightly bound”.

“It’s a fairly nascent market right now,” Kelly said. “Just from our point of view, we see a lot of proof of concepts. And we think we’ll see some of those PoCs make their way commercially into the marketplace.”

At present, there are a handful of use cases across CPS deployments that stand out. Key among them are fault isolation, root cause analysis, and network performance analysis.

“There’re tens of thousands of alarms that come into every Tier-1, Tier-2 operators’ network every day. The fundamental problem is how do you eliminate the noise and really get your teams focused on the root cause of that problem.”

To cut the noise and the gruntwork that comes with it, CSPs are experimenting with agentic AI tools to automate network monitoring, identification of anomalies, and tracking and analysis of root causes. AI agents can observe the network around the clock, isolate issues and flag them as soon as they show up — sometimes even before they show up — and effectively shorten the winding root cause analysis workflow, ensuring faster mean time to repair.

With telco networks becoming more cloud-native and software-dependent, operational complexities have been on the rise recently. “I think this is where this technology can help optimize and maybe provide a little more productivity to the teams that are dealing with a lot of these issues.” That he says will enable them to make services perform better at the user end.

The best opportunity for operators to leverage agentic AI, according to Kelly, is to optimize the mobile radio access network [RAN] where AI can enable cell sleeping for better energy efficiency, predict traffic efficiency for dynamic resource allocation, and help transition from manual and static configuration to self-optimization.

A key takeaway, Kelly noted, is that operators are seeing more success where application is domain-specific. “So rather than trying to go cross domain, looking at a specific area of workflow in their business, bound that, and then really focus on that,” he said. 

According to Appledore’s estimation, the global agentic AI market is set to reach $6.2 billion by 2030, up from $92 million in 2025. Kelly predicts investments to flow into four core areas: digital enablement, service management and operations, network security, and IT and enterprise resource planning (ERP).

As for whether agents could replace humans in that future, Kelly says its unlikely. He said that the narrative is overblown. “I see it more as a productivity enhancer. And I’ll just say, all of the use cases with the operators we’re talking to, they’re still looking at risk mitigation.” Which makes it unlikely for them to rely entirely on agents. 

“I still see human experts in the loop for the foreseeable future,” he concluded.