AT&T Scales Enterprise Agentic AI to 500 Production Workflows
The telecom operator is automating legacy mainframe operations and network provisioning while targeting a threefold expansion next year.

AT&T Corp. has deployed nearly 500 agentic artificial intelligence workflows into active production and plans to triple that footprint over the coming year, according to executive remarks reported by SiliconANGLE .
The telecommunications carrier now processes roughly 45 billion AI tokens daily across the enterprise, with its autonomous agent platform executing over 1 million transactions per month, Luke Corbin, assistant vice president of digital transformation and AI at AT&T Business, said at an Amazon Web Services Inc. event in Boston.
The deployment underscores an enterprise shift from individual employee productivity assistants toward autonomous systems wired directly into operational plumbing. Central to AT&T's rollout is its Business Agent Foundry, an internal group established in early 2025 to re-engineer core operations. Rather than gathering piecemeal automation suggestions, the team maps business processes against standard industry frameworks to excise procedural handoffs and reduce dependencies on aging software stacks.
Technically, AT&T has avoided the prevailing trend of building sprawling fleets of bespoke, task-specific agents. Instead, the company runs a compact set of core agents governed by centralized business logic maintained in context and system memory. Retrieval agents locate pertinent institutional records, while planning agents dynamically determine the sequences, tools, and underlying data needed to execute a workflow.
That infrastructure interfaces directly with AT&T's legacy systems, including mainframe environments. In fiber operations, installation ticket changes previously required technicians to manually retrieve queue items, adjust records, and re-enter them into provisioning pipelines. Automating those mainframe exchanges eliminated manual handoffs and contributed to cutting dedicated internet provisioning times in half for buildings where fiber lines were already present.
To manage operational risk, AT&T separates personal employee tools from production workflows. While staff may spin up productivity bots independently, any workflow touching customer systems or financial operations must undergo security, privacy, and compliance auditing through the Foundry. Workflows initially run with human oversight, transitioning to autonomous execution only after achieving accuracy rates in the mid- to high-90% range within defined risk thresholds.
Corbin claimed that AT&T's production AI initiatives are generating a fivefold return within 12 months across labor savings, operating cost reductions, and automated prospecting, though the company has not published audited financial breakdowns. Looking ahead, AT&T aims to shift its agents from internal maintenance toward launching new commercial services, with an eventual goal of decentralizing the Foundry's oversight directly into operating units.
Sources
Written by
The Company Wire
Inside the companies building what’s next. Reporting on startups, technology, funding and the people shaping them.