AT&T's AI System Slashes Network Outage Downtime by Millions of Hours

United States — Ekhbary News Agency

AT&T has successfully reduced customer downtime by more than 12 million hours through its sophisticated artificial intelligence system, the End-to-End Incident Management (EEIM) platform. This strategic initiative, launched in 2017, aims to identify and resolve network interruptions proactively, ideally before customers even notice, as Andy Markus, AT&T's chief data and AI officer, emphasized.

Proactive Measures Against Costly Disruptions

Keeping its 145 million wireless and 16 million broadband customers connected remains a paramount concern for AT&T. Brief outages, even as short as 20 minutes, can incur substantial costs for the company, often necessitating customer bill credits. A notable incident in February 2024 saw a major wireless outage block over 92 million voice calls and more than 25,000 attempted 911 calls, according to a US Federal Communications Commission report. Such events underscore the critical need for robust network resilience.

Evolving AI for Enhanced Network Management

The EEIM system, initially built on traditional machine learning with technologies from firms like MongoDB and Snowflake, has evolved to incorporate generative and agentic AI capabilities. This expansion allows the tool to pinpoint the root cause of disruptions, prioritize issues, suggest remote fixes, or dispatch human technicians efficiently. It also proactively alerts customers about outages. As it happens, this proactive stance is paying dividends, enabling AT&T to reorganize 10 petabytes of data, including network logs and alarms, to predict and prevent future issues. The strategic investment in advanced AI underscores a critical shift towards predictive maintenance, a move many telecom giants are eyeing.