Leveraging AI to drive CX means continuously gathering data from “real people in real places doing real things,” T-Mobile EVP Ankur Kapoor explained
The telecom industry often frames AI-enabled automation through operational metrics like fewer alarms, faster root-cause analysis, lower mean time to repair and, ultimately, fewer people required to keep increasingly complex systems running. T-Mobile US Executive Vice President and Chief Network Officer Ankur Kapoor starts somewhere else. For the operator, AI-enabled automation is only strategically meaningful when its impact reaches the customer.
“Our vision was no automation is good enough unless it’s really impacting customer experience,” Kapoor said. “If it’s not delivering a differentiated customer experience, it’s only operational benefits.”
That changes how T-Mobile US measures progress. Traditional network KPIs remain necessary, but Kapoor emphasized application responsiveness, successful uploads and video sessions, net promoter scores, network-related calls into customer care and the time required to resolve a customer pain point. While whether a network element moves from red to green on an internal dashboard is certainly relevant to engineers, the operative question is whether the customer can do what they intended to do.
T-Mobile US’ work on the TM Forum-hosted Agentic NOC project illustrates how that principle informs the path toward closed-loop operations. The catalyst, which won an Outstanding Catalyst Award at DTW Ignite 2026, uses specialized agents across network, service, field-force, security and operations domains while retaining human oversight for consequential decisions.
Kapoor described the company’s operating model in three tiers. Tier-one alarm monitoring has been fully automated, allowing machines to identify problems at machine speed. Tier two uses automation for root-cause analysis and fault localization. Tier three involves changes to the live network, where the decision to keep a person in the loop depends on the severity and risk of the action.
The authority to close that loop has to be earned through testing. T-Mobile US compares an agent’s results with the outcome a human operator would have produced, then continuously refines the system. Dynamic CX is one use case where the company is comfortable allowing autonomous action. The platform uses AI alongside a self-organizing network platform to anticipate large-scale events that could change traffic patterns, monitor shifting demand and optimize performance in near-real time.
During this summer’s FIFA World Cup, T-Mobile US said Dynamic CX made more than 11,000 automated optimizations across 12 fan events; its network connected 2.26 million fans across 78 matches while carrying roughly one petabyte of stadium traffic.
This type of intent-based approach to operations introduces a related governance problem: what happens when customer experience, efficiency, security and public-safety requirements compete? Kapoor’s answer is that AI executes policy; it does not create it. “It’s not what governs,” he said. “The business principles are the governance factor in that.”
First-responder communications take precedence, but the system allocates resources according to the actual service need rather than allowing a prioritized user to consume capacity indiscriminately. T-Mobile’s work on conflict management in intent-based networks received the TM Forum Attendees’ Choice Award.
The connective tissue is a knowledge architecture built from real device and application measurements. Kapoor described it as information from “real people in real places doing real things,” analyzed continuously so the network can optimize for the applications customers are actually using rather than for abstract equipment health.
That architecture also links T-Mobile’s domain-level work. The AI-native Live Translation feature turns the core into an AI service platform; AI-RAN pushes intelligence toward the cell site; Dynamic CX and the SON platform convert that intelligence into action.
The architecture may be distributed, but the objective remains singular. T-Mobile is not pursuing agents, AI-RAN or autonomous operations as ends in themselves. As Kapoor put it in closing, “Our vision isn’t just to automate the network.” The point is to enable new services, remove customer pain and make the network respond to what people are actually trying to do.
