- AT&T has a strategy to make its network operate autonomously, with minimal human intervention
- The telco giant is pumping GenAI into more and more aspects of the network lifecycle — all the way down to copper retirement
- But AT&T is implementing GenAI strategically and coinciding with other network transformation activities
From the early stages of development to the final stages of decommissioning, artificial intelligence (AI) is slowly creeping into every corner of AT&T's network lifecycle. The telco giant’s work with generative AI (GenAI) goes well beyond the chatbots operators are famous for — AT&T’s VP of network analytics and automation Raj Savoor told Fierce it is now even using the tech to help with copper retirement efforts.
“We’re taking advantage of both open-source data and commercial GenAI engines to streamline these processes," Savoor said. Traditionally, decommissioning infrastructure would involve sending a technician to assess and document each site. Now, much of this work can be done remotely by using public domain visual analytics data and support systems powered by AI.
The number of manual interventions needed on site has gone down, and some tasks that would have been “left on the table” can be addressed. “Now we can get out of leases of copper poles that we're not using effectively, or there's a better design opportunity,” he added.
The AI systems used across AT&T's network can be anything from basic three-layer models to highly complex neural networks, commercial open source GenAI engines and large language models (LLMs), Savoor explained. AT&T has been using statistical methods and machine learning (ML) for “many years,” which have laid the foundation for these GenAI advancements.
"The opportunity now with GenAI, particularly with embedded deep learning models, is that we can apply these techniques much faster, especially for geospatial data and other visualizations,” Savoor said. "We’re building on what we’ve had. In many cases, we’re not starting from scratch but enhancing existing AI and machine learning capabilities."
AT&T’s path to autonomy
AT&T has been "very opportunistic" in its adoption of GenAI, with use cases numbering in the hundreds. However, the company is strategic about prioritizing implementations that offer the greatest long- and short-term benefits.
The ultimate goal is to achieve an autonomous network — one that can operate with minimal human intervention. Savoor said that the company has been progressing toward this vision for years, with certain areas already demonstrating high levels of autonomy in closed loops (sans human intervention).
That work is happening “at all levels of the network,” he added, particularly with localized autonomous capabilities rather than a “SkyNet” approach.
That includes developing localized capabilities that allow for intelligent decision-making at the node level, such as energy savings.
The company is focused on integrating automation where it makes the most sense, depending on the scale and maturity of the technology.
For instance, as new capabilities like its Cloud RAN are rolled out, they will feature high levels of autonomy right from the start, thanks to advanced local intelligence in disaggregated components, Savoor explained. Meanwhile, existing technologies like LTE, which currently rely on partial automation but still require manual provisioning and configurations, will see more gradual enhancements as they undergo natural upgrade cycles.
As a result, AT&T's gradual GenAI installations will be coincident with other network transformation activities, Savoor noted. "The investment you want to make on legacy, you balance it to say, okay, where do you want to catch the right automation for the right technology scale?" he concluded.