AT&T chief data officer Andy Markus outlined how autonomous agents will become central to the operator’s generative AI (genAI) efforts moving forward, performing tasks such as fraud alerts, network optimisation and software development.

In a blog, the executive stated autonomous agents were known internally as autonomous assistants because they use genAI to assist employees in making decisions or taking specific actions.

He noted large and small language models are designed to harvest information based on prompts. The autonomous assistants then act on that information.

In one application, similar to rival Verizon, it is employing agents to analyse customer accounts when they call in to customer care. Markus stated the agents examine a caller’s account to almost instantly give the staff member a menu of options that can be presented.

A representative for AT&T told Mobile World Live (MWL) the various assistants use different large language models (LLMs) and small language models (SLMs) depending on the task. With the operator interoperable with almost every model, “it’s just about finding the right model to do the task as efficiently as possible”.

The spokesperson stated AT&T currently has 138 autonomous agents and expects that number to increase dramatically next year.

Building master agents
The assistants can also aid AT&T in its software development cycle by assigning specialised agents for specific tasks such as creating user stories, writing code or executing test scripts.

“What’s more exciting, we are exploring how we might be able to use multiple LLM-enabled assistants at the same time to solve more complex business problems,” Markus said. “To do so, we assign one assistant an overall orchestrator persona whose job is to coordinate and reason across other assistant personas with more specialised skills.”

A human developer interacts with a master assistant before it enables additional agents to complete a task.

“Human employees remain responsible for overseeing the assistant’s progress, but this frees up our people to focus on more complex challenges,” Markus stated.

He noted autonomous assistants can take on more tedious tasks in a workflow while producing results faster than a human.

Markus stated AT&T expects autonomous assistants to be deployed more broadly going forward while becoming “integral to helping our employees perform complex operational workflows faster to better serve our customers.”

AT&T was an early adopter of genAI, which could help the operator meet its stated goal of reducing costs by $2 billion over the next few years.