Operators increasingly see AI as a practical tool for network operations. Yet the most useful clues may start with ordinary messages. Research from Srinivasa Maringanti focuses on network fallbacks. These happen when modern services shift to older technologies.
For users, a fallback often looks harmless. An RCS message may become SMS. A VoLTE call may drop to a traditional voice path. The message still arrives, and the call may continue. However, the network has already revealed stress.
“In the case of RCS and fallbacks, or VoLTE calls and fallbacks to traditional systems, it gives us a lot of signals on what’s actually happening in the network,” Maringanti explained.
That signal matters because operators manage enormous data volumes. Billions of messages can hide small but recurring failures. Traditional dashboards often miss these events. They may never trigger alarms, because service continues.
Still, repeated fallbacks can show deeper network instability. They can point to congestion, service mismatch, or local degradation. They can also expose patterns across messaging, voice, and transport systems.
“All these signals are at message level or a call level, it’s humanly impossible to monitor or identify,” Maringanti said.
This is where machine learning can help. Models can compare many small events at scale. They can also link failures across different systems. That may help engineers detect issues before users complain.
However, this approach needs careful preparation. Telecom data often sits in separate silos. Messaging, voice, data, and applications use different systems. Teams may only compare them during major incidents.
“The biggest challenge here is building the integration of these systems together before an LLM even looks into it,” Maranganti said.
Large language models can interpret patterns in context. Yet they need clean, connected data first. Otherwise, operators may spend heavily without gaining useful insight. The model must know what signals deserve attention.
Maringanti also argues for more precise location analysis. City-level averages often hide local problems. A market may look healthy overall. One building, street, or neighborhood may still suffer daily failures.
His research uses geospatial cells, sometimes called hexbins. These divide areas into smaller, trackable sections. At fine levels, they can represent neighborhoods or streets. AI can then compare network behavior within each cell.
“Without AI and ML algorithms, it would be humanly impossible,” he said. “So it’s not a matter of, we can’t go any further [in granularity]. It’s the problem of how to handle such enormous amounts of data.”
The opportunity is clear. AI could turn routine fallbacks into early warning signals. It could help operators improve reliability and user experience. But human expertise remains essential. Networks change constantly, and models can misread unfamiliar events.
For telecom teams, the lesson is practical. Better AI starts with better data context. Sometimes, the first warning of network instability may be a simple text.
