Skip to content
0MESubscribe
SaaS15 Nov 20262 min read

AT&T runs 500 autonomous AI workflows and aims to triple them next year

AT&T is deploying autonomous software agents across legacy systems to automate complex business workflows, cutting installation times and reporting 5x returns.

Listen to this storyAbout 3 min, read by your device

Telecommunications provider AT&T is expanding its use of autonomous artificial intelligence systems to overhaul legacy operations. The company currently operates nearly 500 agentic workflows in production and plans to triple that footprint over the coming year.

Rewiring legacy infrastructure

AT&T handles around 45 billion AI tokens daily, and its agent platform completes over 1 million transactions each month. Rather than merely automating isolated tasks for individual workers, the carrier is using AI agents to redesign entire operating procedures.

To guide this transition, the company formed an internal unit called Business Agent FoundryFoundryA company that makes chips for other companies. Open in glossary in early 2025. The group maps existing operational pathways to remove redundant steps, handoffs between departments, and reliance on aging applications. For instance, the system now automates interactions with legacy mainframe computers. When fiber installation orders change, agents modify records directly instead of human workers manually pulling tickets from queues. In buildings with existing fiber, this process cut internet provisioning times in half.

Central logic and strict oversight

AT&T uses a unified architecture rather than building many standalone bots. A small fleet of agents shares business logic held in context or memory. Retrieval agents collect necessary data, while planning agents organize the actions and tools needed for each job.

Governance rules require that any agent workflow affecting customers or core operations must undergo security, privacy, and compliance checks through the central foundry. Systems run with human supervision until their accuracy hits the mid- to high-90% range within set risk limits. Only then do they operate independently, under continuous tracking for performance drift.

Financial returns and future plans

The company reports fivefold financial returns on its production AI projects within their first year, measured across labor savings, lower operating expenses, revenueRevenueAll the money a company takes in from sales, before costs. Open in glossary generation, and customer satisfaction metrics. Sales teams also use agents to verify customer locations and confirm local network availability before sales reps reach out.

Moving forward, AT&T plans to use agents to enter new markets, introduce services, and support fresh sales channels, with the goal of eventually distributing AI development across the entire enterprise.

How does this story land?

One tap per reader.

Quick poll

Should telecom firms let autonomous AI handle critical customer infrastructure changes?

One tap. No sign-up.

Ask this story

Still have a question?

Answers use only what is in this story. They are written by AI and can be wrong.

Sources

  1. SiliconANGLE, AT&T uses agentic AI to reinvent business processes from the ground up
Companies in this storyAmazon Web ServicesAT&TWords explainedFoundryRevenue
Up next in SaaSCloudflare open-sources Kitesurf, its stateless AI browser for Workers3 min read

Keep reading

All SaaS stories